Lpcox/initial implementation

- Add generic trySDKTransport function to handle common connection logic
- Refactor tryStreamableHTTPTransport to use trySDKTransport
- Refactor trySSETransport to use trySDKTransport
- Reduces code duplication by ~40 lines (Pattern #2 from issue #353)

Co-authored-by: lpcox <15877973+lpcox@users.noreply.github.com>

…5192)

# Test Coverage Improvement: `server.GetServerStatus` /
`server.GetServerIDs`

## Function Analyzed

- **Package**: `internal/server`
- **Function**: `GetServerStatus` (and `GetServerIDs`)
- **File**: `internal/server/unified.go`
- **Previous Coverage**: Partial — the `!state.StartedAt.IsZero()`
branch (running server uptime calculation) was never exercised
- **Complexity**: Medium — iterates server pool, reads concurrent state,
conditional uptime computation

## Why This Function?

`GetServerStatus` has two execution paths:
1. `state.StartedAt.IsZero()` → uptime = 0 (error/not-started state)
2. `!state.StartedAt.IsZero()` → `uptime =
int(time.Since(state.StartedAt).Seconds())` (running state)

The existing test `TestGetServerStatus_ErrorOnFailedLaunch` in
`health_test.go` only covers path #1. Path #2 — where a server
successfully launches and `StartedAt` is set — had zero test coverage.
This branch is core to the health-reporting surface used by agents to
understand server state.

## Tests Added

- ✅ **Empty config** — `GetServerStatus` returns an empty map when no
servers are configured
- ✅ **Running HTTP backend** — covers the `!StartedAt.IsZero()` uptime
branch; asserts `status == "running"` and `uptime >= 0`
- ✅ **Errored server** — stdio server that fails launch reports `status
== "error"` and `uptime == 0`
- ✅ **Mixed-state multiple servers** — one running + one errored,
checked independently
- ✅ **Nil config** — `GetServerStatus` handles nil config without panic
- ✅ **`GetServerIDs` matches config keys** — asserts exact set of server
IDs returned
- ✅ **`GetServerStatus` keys equal `GetServerIDs`** — consistency
invariant between the two methods

## Coverage Report

```
Before: 0% for the running-server uptime branch in GetServerStatus
After:  branch covered via TestGetServerStatus_RunningServer
```

## Implementation Notes

Tests use `httptest.Server` (no Docker required) as a minimal HTTP MCP
backend. The `mockMCPServer` helper in this file responds to
`initialize`, `notifications/initialized`, and `tools/list` JSON-RPC
calls — following the same pattern established in
`unified_http_backend_test.go`. The `decodeJSONRPCMethod` helper is
reused from that file (same package).

---
*Generated by Test Coverage Improver*
*Next run will target the next most complex under-tested function*




> [!WARNING]
> **⚠️ Firewall blocked 1 domain**
>
> The following domain was blocked by the firewall during workflow
execution:
>
> - `proxy.golang.org`
>
> To allow these domains, add them to the `network.allowed` list in your
workflow frontmatter:
>
> ```yaml
> network:
>   allowed:
>     - defaults
>     - "proxy.golang.org"
> ```
>
> See [Network
Configuration](https://github.github.com/gh-aw/reference/network/) for
more information.


> Generated by [Test Coverage
Improver](https://github.com/github/gh-aw-mcpg/actions/runs/25434463369/agentic_workflow)
· ● 18.9M ·
[◷](https://github.com/search?q=repo%3Agithub%2Fgh-aw-mcpg+%22gh-aw-workflow-id%3A+test-coverage-improver%22&type=pullrequests)

<!-- gh-aw-agentic-workflow: Test Coverage Improver, engine: copilot,
version: 1.0.21, model: auto, id: 25434463369, workflow_id:
test-coverage-improver, run:
https://github.com/github/gh-aw-mcpg/actions/runs/25434463369 -->

<!-- gh-aw-workflow-id: test-coverage-improver -->

…etOrLaunchForSession (#7268)

🤖 *This PR was created by [Repo
Assist](https://github.com/github/gh-aw-mcpg/actions/runs/27209796205),
an automated AI assistant.*

## Summary

When `GetOrLaunchForSession` handles an HTTP backend, it previously
called `getServerConfig` once (acquiring the read lock), then
immediately delegated to `GetOrLaunch` which called `getServerConfig` a
second time — acquiring the same read lock twice per HTTP request.

## Root Cause

`GetOrLaunchForSession` needed to know the backend type before routing.
For HTTP backends it called `GetOrLaunch`, which blindly re-fetched the
config:

```go
// Before: two lock acquisitions for HTTP
func GetOrLaunchForSession(...) {
    serverCfg, _ := l.getServerConfig(serverID) // lock #1
    if serverCfg.Type == "http" {
        return GetOrLaunch(l, serverID)          // lock #2 inside
    }
}
```

## Fix

Extract the inner logic of `GetOrLaunch` into a private
`getOrLaunchWithConfig` helper that accepts a pre-fetched
`*config.ServerConfig`. Both public functions now call this helper after
their single config lookup:

```go
// After: one lock acquisition per request
func GetOrLaunchForSession(...) {
    serverCfg, _ := l.getServerConfig(serverID)  // lock #1 only
    if serverCfg.Type == "http" {
        return getOrLaunchWithConfig(l, serverID, serverCfg) // no extra lock
    }
}
```

## Trade-offs

- **No public API changes** — `GetOrLaunch` and `GetOrLaunchForSession`
signatures are unchanged.
- **No behaviour change** — The new `getOrLaunchWithConfig` is a
verbatim extraction of the existing `syncutil.GetOrCreate` call body.
- **Minor lock contention reduction** — For HTTP backends routed through
`GetOrLaunchForSession`, the `getServerConfig` RLock is acquired once
instead of twice.

Addresses: #7254

## Test Status

⚠️ **Infrastructure failure**: `go build` and `go test` require
downloading modules from `proxy.golang.org`, which is blocked by the
network firewall in this environment. The code change is a
straightforward extraction with no behavioural changes — the
`getOrLaunchWithConfig` body is an exact verbatim copy of the original
`syncutil.GetOrCreate` closure in `GetOrLaunch`. All existing tests for
`GetOrLaunchForSession_HTTPBackend`, `GetOrLaunch_*`, and related
launchers remain valid and unchanged.

- ✅ `gofmt` — no formatting changes
- ⚠️ `go build` — blocked (proxy.golang.org unreachable in CI sandbox)
- ⚠️ `go test` — blocked (same reason)




> [!WARNING]
> <details>
> <summary>Firewall blocked 1 domain</summary>
>
> The following domain was blocked by the firewall during workflow
execution:
>
> - `proxy.golang.org`
>> To allow these domains, add them to the `network.allowed` list in
your workflow frontmatter:
>
> ```yaml
> network:
>   allowed:
>     - defaults
>     - "proxy.golang.org"
> ```
>
> See [Network
Configuration](https://github.github.com/gh-aw/reference/network/) for
more information.
>
> </details>


> Generated by [Repo
Assist](https://github.com/github/gh-aw-mcpg/actions/runs/27209796205) ·
sonnet46 5.2M ·
[◷](https://github.com/search?q=repo%3Agithub%2Fgh-aw-mcpg+%22gh-aw-workflow-id%3A+repo-assist%22&type=pullrequests)
>
<details>
<summary>Add this agentic workflows to your repo</summary>

To install this agentic workflow, run

```
gh aw add githubnext/agentics@851905c
```
</details>


<!-- gh-aw-agentic-workflow: Repo Assist, engine: copilot, version:
1.0.52, model: claude-sonnet-4.6, id: 27209796205, workflow_id:
repo-assist, run:
https://github.com/github/gh-aw-mcpg/actions/runs/27209796205 -->

<!-- gh-aw-workflow-id: repo-assist -->