`PushBuffers::clear_ranges` is quadratic and leaks

**Describe the bug**

`PushBuffers::clear_ranges` performs an O(N*M) scan to release consumed buffers, where N is the number of buffered ranges and M is the number of ranges to clear. On wide schemas (10k+ columns), this produces quadratic overhead in `PushDecoder` row group construction.

Additionally, `clear_ranges` matches buffers by exact range equality. When the IO layer coalesces adjacent requested ranges into fewer, larger fetches, the coalesced buffer never exactly matches any individual requested range, so `clear_ranges` silently skips it. The buffer leaks in `PushBuffers` until the decoder finishes or the caller manually calls `release_all_ranges`, increasing peak RSS proportionally to the amount of data coalesced ahead of the current row group.

This puts coalescing in a bind: without it, buffer count scales with range count and the quadratic `clear_ranges` dominates. With it, memory is not reclaimed incrementally.

**To Reproduce**

Use `PushDecoder` on a Parquet file with a wide schema (10k+ columns). Push data without coalesced buffers. Observe row group construction time scaling quadratically with column count. Conversely if using coalescing, observe memory not being released by `clear_ranges`, growing RSS until the decoder finishes.

**Expected behavior**

Buffer release should scale with buffer count (not range count), and coalesced or arbitrarily-sized buffers should be released incrementally as the decoder progresses.

**Additional context**

Start with PushBuffers::clear_ranges and trace how PushDecoder constructs row groups and tracks buffered ranges. Reproduce the wide-schema and coalesced-buffer cases, then verify that release work scales with buffer count and that coalesced buffers are reclaimed incrementally.

Written by the indexing model from the issue text.