fix(proxy): thread-anchor guard on threads responses + honest tool:response success

[debugmcpdev]

Fixes #396
Fixes #397

Two small correctness/telemetry fixes from the attach-hardening follow-ups.

#396 — ProxyManager no longer clobbers currentThreadId with threads[0]

handleDapResponse captured a thread id from every successful threads response, unconditionally overwriting the id anchored by a stopped event. Any list_threads call (or internal threads poll — the attach verify loop, the stack-readiness scan, pause discovery) landing while paused on a non-first thread silently retargeted subsequent stackTrace/scopes/evaluate to threads[0].

The capture now only fires when no thread id is known (currentThreadId == null), preserving the fallback for adapters that omit threadId from stopped events.

• New test: stopped event anchors thread 12 → a threads response listing [1, 12] no longer retargets to 1.
• The existing fallback-capture test (no prior anchor → captures threads[0]) still passes unchanged.
• The dap-core functional mirror only captures on stopped already, so no core divergence.

#397 — tool:response log line reports the payload's own success

The structured tool:response line hardcoded success: true (meaning "the handler didn't throw"), directly contradicting payloads like a failed attach_to_process that return { success: false } without throwing — actively misleading during triage and hiding failure rates from log-based metrics.

extractPayloadSuccess() now parses the first text content entry and mirrors its boolean success when present; payloads without one (or non-JSON) still log success: true. tool:error (thrown path) is unchanged.

• New tests assert tool:response carries success: false for a failing attach payload and success: true for a successful one.
• docs/logging-format-specification.md updated: documents the new success semantics and drops the response field that was never actually emitted.

Verification

• npx vitest run tests/unit/proxy/ tests/core/unit/server/ — 37 files, 663 tests pass
• Full npm test green
• npm run lint clean

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 90.00000% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/server.ts	88.88%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!