Fix memory leak by replacing @lru_cache with instance-level caching in data parsers

[detecti1]

Summary

Replace @lru_cache on instance methods with instance-level caching to fix memory leak. The global LRU cache keys by self, retaining parser instances and their DataFrames indefinitely.

What's Changed

• BaseDataFrameDataParser: Add _field_metas_cache, _raw_fields_cache, _cache_lock for thread-safe instance-level caching with double-checked locking.
• SparkDataFrameDataParser / DatabaseDataParser / CloudDatasetParser: Apply the same pattern.
• Keep @lru_cache only for pure function get_timezone_base_offset.

Behavior Compatibility

1. Public API unchanged (same properties and return structures).
2. Semantics preserved: once-per-instance computation, reuse for subsequent accesses.
3. @lru_cache automatically adds cache_info() and cache_clear() methods to decorated functions. This project does not use these methods internally, but external code that explicitly calls them will raise AttributeError after this change (e.g., BaseDataFrameDataParser.field_metas.fget.cache_clear()).

Breaking Changes

None for normal usage. Only affects code introspecting or manipulating the LRU cache internals.

Related Issues

Fixes #723

[ObservedObserver]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Nice work!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".