Replace dynamic evaluation in filtering

[YT-er]

Summary

• replace dynamic eval() use in tabular filtering with an explicit operator map
• parse filter values and thresholds as numeric input
• reject unsupported operators and malformed values with contextual errors
• add regression coverage for every supported comparison, invalid operators, and malicious input

Why

Filter values originate in TSV data. Constructing and evaluating Python expressions from those values can execute unintended code and produces unclear failures for malformed input.

Impact

Valid numeric filters retain their existing behavior, including NA, skipped values, and inf. Invalid operators and non-numeric values now fail deterministically.

Validation

• python -m unittest tests.test_filter — 13 tests passed
• Python compile check passed for the changed source and test
• git diff --check passed

This is one focused replacement for part of #1430 and is based on 8.0.0.

AI assistance disclosure

This PR was AI-assisted using OpenAI Codex for implementation, test drafting, and command execution. I reviewed the diff line by line, ran the validation above, and take responsibility for the submitted changes.