feat: harden input parsing with size caps, depth limits, and filesystem validation

[Arkptz]

Summary

Add defense-in-depth security caps to prevent resource exhaustion and filesystem-based attacks:

• Cap TNetString payload size at 256 MiB and recursion depth at 256 levels
• Cap JSON-to-schema conversion depth at 64 levels
• Replace hand-rolled glob matcher with battle-tested globset crate
• Reject symlinks, FIFOs, and non-regular files by default (with --allow-symlinks escape hatch)
• Validate input file size against configurable --max-input-size limit
• Validate path parameter identifiers against safe character set
• Bound HAR format-detection read to 4 KiB instead of reading entire file
• Expose --max-input-size, --max-payload-size, --max-depth, --max-body-size CLI flags

Type of Change

• Bug fix (non-breaking)
• New feature (non-breaking)
• Breaking change (fix or feature that would cause existing behavior to change)
• Documentation update
• Refactoring / internal cleanup
• CI / build / tooling

Checklist

• cargo fmt --all clean
• cargo clippy --all-targets --all-features -- -D warnings clean
• cargo test passes locally
• Added or updated tests for the change
• Updated README / CLI help text if user-facing behavior changed
• Added entry to CHANGELOG.md under [Unreleased] (if user-facing)
• Used conventional commit style in commit messages (feat:, fix:, docs:, etc.)

Testing

• Unit tests added for TNetString payload cap and depth limit
• Unit test for JSON schema depth limit
• Unit tests for path parameter identifier validation
• Integration test suite (tests/security.rs) covering symlink rejection, FIFO rejection, oversize input rejection, and symlink opt-in
• All 192 lib tests + 5 security integration tests pass
• cargo publish --dry-run succeeds

Related Issues

First phase of input hardening. Error::BodyTooLarge variant is defined but reserved for a future PR.