…nd-trip

Merge #4999

## Summary

Fixes #4801 — Adaptive thinking is broken when using Claude models through LiteLLM.

## Root Cause

When Claude produces extended thinking with `thinking_blocks` (each containing a `type`, `thinking` text, and `signature`), the round-trip through ADK's LiteLLM integration silently loses them:

1. **`_extract_reasoning_value()`** only read `reasoning_content` (a flattened string without signatures), ignoring the richer `thinking_blocks` field
2. **`_content_to_message_param()`** set `reasoning_content` on the outgoing `ChatCompletionAssistantMessage`, but LiteLLM's `anthropic_messages_pt()` prompt template **silently drops** the `reasoning_content` field entirely
3. Result: thinking blocks vanish from conversation history after turn 1; Claude stops producing them

## Fix

Three coordinated changes in `lite_llm.py`:

| Change | What it does |
|---|---|
| `_is_anthropic_provider()` helper | Detects `anthropic`, `bedrock`, `vertex_ai` providers |
| `_extract_reasoning_value()` | Now prefers `thinking_blocks` (with per-block signatures) over `reasoning_content` |
| `_convert_reasoning_value_to_parts()` | Handles `ChatCompletionThinkingBlock` dicts, preserving `thought_signature` |
| `_content_to_message_param()` | For Anthropic providers, embeds thinking blocks directly in the message `content` list as `{"type": "thinking", ...}` dicts — this format passes through LiteLLM's `anthropic_messages_pt()` correctly |

For non-Anthropic providers (OpenAI, etc.), behavior is unchanged — `reasoning_content` is still used.

## Verification

- LiteLLM's `anthropic_messages_pt()` was tested to confirm:
  - `reasoning_content` field → **DROPPED** (existing LiteLLM bug)
  - `content` as list with `{"type": "thinking", ...}` → **PRESERVED** ✅
  - Signatures in thinking blocks → **PRESERVED** when in content list ✅

## Tests

Added 7 targeted tests covering:
- `_is_anthropic_provider()` — provider detection
- `_extract_reasoning_value()` — prefers `thinking_blocks` over `reasoning_content`
- `_convert_reasoning_value_to_parts()` — signature preservation from block dicts
- `_convert_reasoning_value_to_parts()` — plain string fallback (no signature)
- `_content_to_message_param()` — Anthropic: thinking blocks embedded in content list
- `_content_to_message_param()` — OpenAI: reasoning_content field used (unchanged)
- `_content_to_message_param()` — Anthropic thinking + tool calls combined

Full test suite: **4732 passed, 0 failures**

> ⚠️ This reopens #4811 which was accidentally closed due to fork deletion.

Co-authored-by: George Weale <gweale@google.com>
COPYBARA_INTEGRATE_REVIEW=#4999 from giulio-leone:fix/litellm-anthropic-thinking-roundtrip 26bf73a
PiperOrigin-RevId: 938711589