Replace panicking .unwrap()/.expect() with error propagation in production code

[Copilot]

Several production code paths could crash the daemon via unrecoverable panics — lock poisoning cascades, unchecked HTTP header construction, and finish_batch() schema assertions that swallow the error type entirely.

Changes

Lock poisoning (enrichment.rs)

5× .expect("*lock poisoned") on RwLock replaced with .unwrap_or_else(|e| e.into_inner()). A panicking thread no longer poisons K8sPathTable, CsvFileTable, and JsonLinesFileTable for all subsequent readers/writers.

HTTP header construction (diagnostics.rs, main.rs)

6× .unwrap() on tiny_http::Header::from_bytes() replaced with map_err(|()| io::Error::other(...))? in functions that already return Result.

Builder finish_batch() (storage_builder.rs, streaming_builder.rs)

Return type changed from RecordBatch to Result<RecordBatch, ArrowError>; .expect() replaced with ?.

Scanner scan() (scanner.rs)

Both SimdScanner::scan() and StreamingSimdScanner::scan() now return Result<RecordBatch, ArrowError>. UTF-8 validation failures return Err(ArrowError::InvalidArgumentError(...)) instead of panicking:

// Before
std::str::from_utf8(buf).expect("SimdScanner: input is not valid UTF-8");

// After
std::str::from_utf8(buf).map_err(|e| {
    ArrowError::InvalidArgumentError(format!("SimdScanner: input is not valid UTF-8: {e}"))
})?;

#[should_panic] tests converted to assert!(result.is_err()).

Pipeline error handling (pipeline.rs)

Scan stage now handles Result from scan(), returning io::Error on failure rather than unwinding.

Callsite updates

All bench, example, fuzz-target, and integration-test callers of the changed APIs updated to .expect() / .unwrap() as appropriate for non-production code.

---

💬 Send tasks to Copilot coding agent from Slack and Teams to turn conversations into code. Copilot posts an update in your thread when it's finished.