expr: don't read unordered stats bounds as an empty range

[def-]

ResultSpec::value_between collapsed min > max to nothing(), i.e. "no value can be here". That only holds for bounds a caller derived coherently. Persist part statistics are not such a caller: arrow orders floats totally, putting -NaN below -Infinity, while OrderedFloat, the Datum order value_between compares in, ranks every NaN above every finite value. A part holding -NaN therefore reports lower = -NaN against a finite upper, the range read back as empty, and filter pushdown discarded the part, losing every other row in it. '-NaN' is ordinary user input, so no corrupt storage is needed to reach this.

Widening unordered bounds in value_between is not sufficient on its own. A part holding NaNs of both signs records the total-order bounds (-NaN, +NaN), which decode under OrderedFloat to NaN == NaN: a seemingly valid, non-inverted range claiming the part holds nothing but NaN. A filter that a finite row matches but NaN does not (for example value = 0.0) still discarded the part. By the time value_between runs, OrderedFloat has erased the NaN signs, and (NaN, NaN) is also the legitimate spec of a genuinely all-NaN part, so the guard has to live in the stats decode, which still sees the sign bits: mz_repr::stats::col_values now refuses to produce float bounds from a negative-NaN lower, unless the upper shows the whole column is NaN.

Tests: unit regressions for both symptoms in persist_source.rs's filter_pushdown_audit module (negative_nan_does_not_discard_matching_part, mixed_sign_nans_do_not_discard_matching_part, each verified to fail without its half of the fix), and peek + dataflow + EXPLAIN FILTER PUSHDOWN cases for both shapes in test/sqllogictest/explain/pushdown.slt.

Closes: PER-53

🤖 Generated with Claude Code

[def-]

Superseded by #38108 (same commit, head branch in this repo so it can take part in a GitHub stack). Bottom of stack #38111: #38108 → #38109 → #38110.