perf(fspy): use sparse Windows shared memory

[wan9chi]

Motivation

Avoid allocating a fully backed temporary file for each tracked Windows task while keeping file-access recording on the shared-memory fast path.

Background

The Windows adapter relied on shared_memory for mapping creation, naming, and cleanup. Owning the native mapping lets fspy share it by kernel object name, keep sender opens independent of the temporary directory, and mark the backing file sparse.

Windows may write dirty mapped pages under memory pressure. Record writers still use atomics and pointer writes without a file or IPC system call for each record.

Changes

• Replaces the Windows shared_memory adapter with a small windows-sys implementation and an owned mapped-view guard.
• Uses UUID identifiers and derives the named mapping from the identifier and expected size.
• Creates a temporary delete-on-close backing file and marks it sparse before extending it.
• Treats unavailable sparse-file support as a creation error instead of falling back to full allocation.
• Opens sender views by mapping name without reopening the backing file or depending on the sender's TMP, TEMP, or working directory.
• Uses windows-sys imports for LocalFree and PathAllocCombine, replacing linker inputs that the removed dependency supplied.
• Covers cross-process sharing, owner cleanup, size validation, sparse allocation, changed environments, and production-size endpoint access.

[wan9chi]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• perf(fspy): use sparse Windows shared memory #524 : 2 dependent PRs (#525 , #526 ) 👈 (View in Graphite)
• fix(fspy): use memfd for Linux shared memory #523
• test(e2e): reproduce constrained dev shm failure #522
• docs(fspy): document shared-memory facade ownership semantics #521
• refactor(fspy): introduce fspy_shm facade #520
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: adc5221ce4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".