feat: release unused python memory

[PeterStaar-IBM]

Summary

This change introduces a configurable native memory trimming policy to reduce allocator fragmentation during page processing.

resolving: #273, #227

What changed

• Added release_native_memory_every_n_pages to DecodePageConfig
  • 0 disables trimming
  • N > 0 triggers trimming every N processed pages
• Added a shared cross-platform helper for native memory release:
  • glibc: malloc_trim(0)
  • macOS: malloc_zone_pressure_relief(nullptr, 0)
  • Windows: HeapCompact(GetProcessHeap(), 0)
• Added an internal atomic page counter to the threaded processing base and invoke trimming after completed page results
  are produced
• Applied the same throttled trimming policy to the non-threaded docling_parser path, using the decode_config passed to
  get_page_decoder(...)
• Exposed the new config field through pybind and propagated it through the Python config copy helper

Design notes

• Trimming is now driven by decode-time configuration instead of unconditional calls in unload paths
• The policy is shared across:
  • threaded parser
  • threaded renderer
  • non-threaded parser
• The setting is named for native memory, not Python memory, since this operates on the process allocator rather than
  Python GC

Expected impact

• Reduces long-run native heap fragmentation during large decode workloads
• Avoids trimming on every page by making the cadence configurable
• Keeps the feature opt-in by default via 0 disabled

Example

cfg = DecodePageConfig()
cfg.release_native_memory_every_n_pages = 50

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[github-actions]

✅ DCO Check Passed

Thanks @PeterStaar-IBM, all your commits are properly signed off. 🎉

[simonschoe]

@PeterStaar-IBM awesome, really appreciate that your are relentlessly working on the core issue. Kudos to you and the team! 🫶