docs: add debugging tips to DEVELOPMENT.md

[alliasgher]

What this PR does

Expands DEVELOPMENT.md with a Debugging tips section under "How to run the controller locally", covering the items listed in #666:

• RUNTIME_NAMESPACE — restrict watched namespaces to reduce noise
• --concurrent=1 — serialise reconcile loops for easier tracing
• HOSTNAME — set controller identity for local leader election
• Suspending irrelevant objects — patch spec.suspend: true on unrelated resources
• Scaling down the in-cluster controller — avoid competing reconciliations

Fixes #666.

Assisted-by: Claude/claude-opus-4-7

[SyedAsadRazaDevops]

Small issue with the make run ARGS=--concurrent=1 example: the current run target does not pass ARGS through to go run, so --concurrent=1 would not reach the controller.

Could this either use a direct go run ... --concurrent=1 example that preserves the existing --storage-adv-addr=:0 and --storage-path=$(PWD)/bin/data flags, or update the Makefile to support extra run args?

[alliasgher]

Good catch, fixed in a1ba6ca using a direct go run invocation with the run target's storage flags.

[alliasgher]

Closing this. Two of the three tips were wrong rather than merely imprecise, and once they are removed there is not enough left to justify a review cycle.

For the record, so this is searchable if anyone tries the same thing later:

• Limit the watched namespace was wrong. RUNTIME_NAMESPACE is only read when --watch-all-namespaces is false, and that flag defaults to true, so the documented command had no effect. @stefanprodan also made the stronger point that narrowing the watch would hide cross-namespace behaviour you often need while debugging, so the section should not exist even in a corrected form.
• Reduce reconcile concurrency was backwards as general advice: goroutine leaks, deadlocks and races only surface under concurrency, and it is redundant once unrelated objects are suspended.
• Suspend unrelated objects was the only sound tip, and it amounts to mentioning that flux suspend <kind> --all exists, which is not worth a PR on its own.

Thanks @stefanprodan for the review that caught this, and apologies to the other repos for receiving the same wrong text. If a maintainer does want a debugging section in DEVELOPMENT.md, I am happy to write one from scratch against what the flags actually do.