Optimize parsing, filtering, and rendering performance

[Marc-André Moreau (mamoreau-devolutions)]

Summary

This PR reduces allocation and CPU overhead across parsing, filtering, query execution, and table rendering in cirup_core.

Highlights:

• Replace regex-backed text filter compilation and matching with a purpose-built matcher for the supported pattern subset.
• Replace prettytable-rs table rendering with a manual ASCII formatter while preserving the existing output layout.
• Remove dot_json-based JSON flattening in favor of direct recursive flattening.
• Replace the RESTEXT regex parser with a manual parser and direct writer.
• Reduce RESX escaping and parsing allocations by writing escapes directly into output buffers and iterating parsed nodes directly.
• Reuse fast in-memory query execution for rusqlite-backed queries and avoid extra row cloning when reading DB results.
• Preallocate file reads and avoid redundant UTF-8 output copies when writing resource files.

Benchmark Highlights

• Query filter compilation (5,000 key patterns + 5,000 value patterns): ~1.251s -> ~2.451ms
• Filter pass on 50,000 resources with key and value filters: ~2.57ms -> ~1.71ms
• Repeated resource table rendering (20,000 iterations): ~102.6ms -> ~13.3ms
• Repeated report table rendering (20,000 iterations): ~193.1ms -> ~26.5ms

Validation

• cargo fmt --all
• cargo test --workspace
• cargo clippy --workspace --all-targets