fix(core): reset attempt timeout on received chunks (#87)

[leon-zym]

Summary

Fixes #87.

In streamCommandCode, attemptTimeoutId was started when the HTTP request begins, but was not refreshed inside readLoop when reader.read() yielded new chunks. This caused timeoutMs (forwarded from Pi's httpIdleTimeoutMs setting, default 300,000ms) to act as a hard wall-clock cap on the total duration of the generation rather than an idle gap timeout between chunks, aborting legitimate long-running reasoning generations (e.g. GLM-5.3-flash with max thinking or DeepSeek models) after 300 seconds.

Changes

• In src/core.ts (readLoop):
  • Clear the timeout when done is true (response completed).
  • Refresh attemptTimeoutId whenever a new chunk is received, preserving the idle timeout protection against hanging streams while allowing active generations of any length.
• In tests/test-retry.ts:
  • Added regression test does not abort actively streaming responses that exceed timeoutMs overall (streams for 90ms with 30ms gaps under a 50ms timeout).
• In CHANGELOG.md:
  • Added Unreleased entry referencing Long reasoning streams abort after 300s: treat timeoutMs as chunk idle timeout instead of total wall-clock limit #87.

Test plan

• npm run typecheck — 0 errors
• npm run format:check — Prettier clean
• npm run test:unit — all unit tests pass, including the new regression test and all existing retry/timeout tests

[patlux]

Integrated current main and widened the active-stream regression timing margin (200ms gaps, 500ms idle limit, >600ms total). Confirmed this test fails on the original implementation and passes with the fix. Existing regression explicitly verifies a stream that emits partial text and then stalls times out without retrying; it remains green. Full npm test passes with real Pi and OMP plus format/diff checks. Waiting for CI.

[patlux]

Shipped in pi-commandcode-provider@0.7.0: https://github.com/patlux/pi-commandcode-provider/releases/tag/v0.7.0 . Public npm package model discovery and GOAT live smoke verified.