perf(query): enforce limit parameter and reduce compact row parsing allocations

[ZhuchkaTriplesix]

Summary

QueryParams.limit (sent by Querya Desktop as AppSettings.sqlResultMaxRows, typically 1,000–5,000) was deserialized in src/rpc/handlers/query.rs but never used anywhere. db.query always parsed and materialized the entire result set, which risks OOM on the 256 MB ClickHouse Sandbox ceiling for large tables.

• parse_compact_output (src/mapper/row_compact.rs) now takes an Option<usize> limit and stops parsing/allocating further data rows once it's reached, instead of parsing everything and discarding the excess. handle_query passes QueryParams.limit through on both the mock and real ClickHouse paths.
• Replaced the Vec<&str> line collection with direct iteration over output_text.lines(), dropping the intermediate allocation of the whole line list before any parsing starts.
• Added query_id directly to QueryResult instead of round-tripping through serde_json::to_value(...) once just to mutate the resulting Value via as_object_mut() to splice in a queryId key.

All other parse_compact_output call sites (schema introspection in schema.rs, tree building in tree.rs) pass None — they read small, bounded system tables and don't need truncation.

Closes #47

Test plan

• cargo test — 98/98 passing, including new regression tests: parse_compact_output limit enforcement (under/over/zero/no limit) and its query_id field, plus an end-to-end handle_query mock test asserting a limit: 1 request actually returns 1 row
• cargo fmt --all -- --check
• cargo clippy --all-targets --all-features -- -D warnings