DiskSeismic: add python tests & de-dup serialization

[zirui-song-18]

Description

Addresses the two follow-ups from #32:

1. De-duplicate the serialization. DiskSeismicIndex wrote the per-cluster doc-id membership twice — in the posting-list section and as per-block doc ids in the inline forward index — yet the posting-list copy is never read on the mmap search path (doc ids come from the forward index; only the cluster summaries + counts are used). write_index now serializes the posting lists summaries-only, emitting docs_/offsets_ as count-0 arrays. Since the align helpers round-trip empty arrays byte-for-byte, the reader is unchanged and plain SeismicIndex output stays byte-identical.
2. Python integration tests. Adds python_tests/test_disk_seismic_index.py (mmap-only contract, top-k′ block budget monotone/saturating, bit-exact fresh-build vs mmap-reload, id-map, filtering, empty/unbuilt, padding). No binding changes needed — disk_seismic is already reachable through SWIG.

Issues Resolved

• RFC: [RFC] Disk-resident forward index for SEISMIC, with an inline on-disk layout #21

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[chishui]

does the index size results in previous PR change much?

[zirui-song-18]

does the index size results in previous PR change much?

Thanks for your question. Following is a benchmark:

• lambda 4000: -0.3854% compared with 439b992
• lambda 6000: -0.4052% compared with 439b992