fix: async cache_clear() TypeError + nested numpy serialization

[27Bslash6]

Summary

Combined fix for two related serialization/decorator bugs.

Fix 1: cache_clear() broken for async functions (#49)

cache_clear() called ainvalidate_cache() without awaiting, silently discarding the coroutine — cache was never cleared. Now raises TypeError directing users to await fn.ainvalidate_cache().

Root cause: wrapper.py:1395 — sync function cannot await a coroutine.

Fix 2: serializer="auto" fails on nested numpy arrays (#50)

_auto_default() had no handler for ndarray, so msgpack.packb() raised TypeError when numpy arrays were nested inside dicts or lists (despite documentation claiming support). Added __ndarray__ marker to _auto_default() and _auto_object_hook() for roundtrip serialization.

Root cause: Top-level numpy detection only; msgpack fallback path had no numpy handler.

Test plan

Issue #49 (4 tests)

• test_cache_clear_raises_type_error_for_async_function
• test_cache_clear_error_message_suggests_ainvalidate
• test_cache_clear_does_not_raise_for_sync_function (regression guard)
• test_async_ainvalidate_cache_still_works

Issue #50 (19 tests)

• Dict with 1D/2D/float64/multiple numpy arrays
• Lists containing numpy arrays
• Deeply nested structures, mixed types (numpy + datetime + UUID + set)
• Edge cases: empty array, 0-d scalar, 3D tensor
• With and without ByteStorage integrity checking
• Corruption detection: malformed __ndarray__ markers

Closes #49
Closes #50

[codecov]

Codecov Report

❌ Patch coverage is 0% with 11 lines in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/cachekit/serializers/auto_serializer.py	0.00%	7 Missing and 1 partial ⚠️
src/cachekit/decorators/wrapper.py	0.00%	3 Missing ⚠️

📢 Thoughts on this report? Let us know!