fix(core): resolve throttling race condition and optimize memory cache expiration

[AtaCanYmc]

Description

This PR addresses critical concurrency, context management, and performance issues:

1. Thread-Safety in LocalThrottler (githubkit/throttling.py):

   • Added threading.Lock to guard lazy initialization of _semaphore and _async_semaphore.
   • Resolves a race condition where concurrent threads could instantiate duplicate semaphores and bypass the max_concurrency limit.

2. Harden Context Manager Teardown in GitHubCore (githubkit/core.py):

   • Updated __exit__ and __aexit__ to safely check for non-None client instances before calling .close() / .aclose().
   • Prevents potential AttributeError crashes during exception teardown.

3. $O(1)$ Passive Cache Expiration in MemCache (githubkit/cache/mem_cache.py):

   • Replaced full $O(N)$ dictionary scans on every read/write with passive $O(1)$ item expiration during get() lookups.

4. Unit Test Coverage (tests/test_unit_test/test_unit_test.py):

   • Added unit test coverage validating thread-safety under multi-threaded execution, passive cache expiration, and context manager cleanup.

Verification

Ran the unit test suite:

uv run pytest tests/test_unit_test