geotiff: cover sparse-strip parallel decode in _read_strips and HTTP path (#2100)

[brendancol]

Summary

The strip-decode parallelisation in #2100 / #2104 added a collect-decode-place pipeline in both _read_strips and _fetch_decode_cog_http_strips. The job-collection loop filters sparse strips (byte_counts[idx] == 0) before they reach the ThreadPoolExecutor. test_parallel_strip_decode_2100.py covers parallel/serial parity, the pool-engaged branch, single-strip short-circuit, windowed reads, and planar=2, but every fixture is fully populated. The 128x128 sparse fixture in test_sparse_cog.py is below the 64K-pixel parallel gate, so the sparse-strip filter inside the parallel branch is untested.

A regression that lost the byte_counts==0 guard would silently ship: the decoder would receive an empty data slice and either raise Decompressed tile/strip size mismatch or return corrupt pixels.

Seven new tests cover the gap:

• local-strip full-image parallel/serial parity with sparse strips
• parallel-pool-engaged on a multi-strip sparse image
• windowed read across the sparse boundary
• all-sparse degenerate input
• planar=2 multi-band sparse parity
• HTTP windowed read on a non-sparse strict subset
• HTTP windowed read across the sparse boundary

Mutation against the strip-job collection sparse guard (delete the byte_counts == 0 continue) flips 5 of 5 local tests red with Decompressed tile/strip size mismatch: expected ... got 0; mutation against the HTTP path sparse guard at line 2646 flips the boundary HTTP test red. Source untouched.

Test plan

• pytest xrspatial/geotiff/tests/test_parallel_strip_decode_sparse_2100.py -v (7 / 7 pass locally)
• pytest xrspatial/geotiff/tests/test_parallel_strip_decode_2100.py xrspatial/geotiff/tests/test_parallel_strip_decode_sparse_2100.py (16 / 16 pass)
• Mutation: drop sparse guards in _read_strips, confirm new tests fail
• Source restored to baseline (md5sum matches pre-mutation copy)

deep-sweep-test-coverage pass 18; HIGH category 3 + 4.