Potential unsoundness in filter_run_end_array

[cramertj]

I believe this comment is incorrect:

arrow-rs/arrow-select/src/filter.rs

Lines 444 to 447 in b945b15

	// in filter_array the predicate array is checked to have the same len as the run end array
	// this means the largest value in the run_ends is == to pred.len()
	// so we're always within bounds when calling value_unchecked
	for pred in (start..end).map(|i| unsafe { filter_values.value_unchecked(i as usize) }) {

It allows for re_arr.len() <= pred.len().
Instead, filter_array only errors in the case where values.len() is smaller, not where it is larger.

arrow-rs/arrow-select/src/filter.rs

Line 341 in b945b15

if predicate.filter.len() > values.len() {

Furtnermore, since this is a requirement placed on callers of filter_run_end_array, filter_run_end_array should be changed to unsafe fn and include a comment like the following:

/// # Safety
///
/// The caller must ensure that the `pred` and `re_arr` are the same length.

[alamb]

Sounds like a good thing to fix. Thank you 🙏

[delamarch3]

take

[alamb]

label_issue.py automatically added labels {'arrow'} from #6675