hid: fix bug in report parsing when reports are not declared contiguously

[williampMSFT]

#933 had a bug: it was not correctly handling the case where a report ID gets 'revisited', but it turns out it's legal to go back to an earlier report ID and keep adding to it, e.g.

• Report ID = 1
• declare 2 bytes
• Report ID = 2
• declare 3 bytes
• Report ID = 1
• declare 2 bytes

means that report ID 1 is a 4-byte report. This changes our parsing logic to account for this; we have to track each individual report instead of just the largest observed so far.

[Copilot]

Pull request overview

Summary of changes

This PR fixes a correctness bug in HID report-descriptor parsing where report fields for the same Report ID can appear in multiple non-contiguous regions of the descriptor. Instead of treating “switching report IDs” as ending the current report and only tracking a running max, the parser now accumulates bit sizes per Report ID across the entire descriptor. This aligns the implementation with legal HID descriptor patterns where a report ID can be revisited and extended later. New unit tests cover both non-contiguous Report ID re-entry and Push/Pop restoring a prior report ID.

Changes:

• Track accumulated Input/Output/Feature bit counts per Report ID while parsing, then compute the max at the end.
• Remove the previous “flush on report-id switch / pop-to-different-id” logic that incorrectly split reports.
• Add tests for non-contiguous Report ID declarations and for Report ID restoration via Pop.

Step-by-step review guide

1. Parsing logic change: accumulate per Report ID

   • The constructor now maintains a reports table keyed by state.report_id, and adds each Main item’s computed bits to the appropriate entry.
   • This matters because HID descriptors can legally interleave report IDs; the final report size must reflect the sum of all fields for that ID, regardless of contiguity.

2. Push/Pop semantics

   • Pop now simply restores the prior GlobalItemState without forcing a “report boundary.”
   • This is important because Push/Pop can restore a prior report ID and continue adding fields to the same report, which should be accumulated rather than treated as a new report.

3. Final max computation

   • After parsing, the code iterates across all tracked reports and computes the maximum input/output/feature payload sizes.
   • This preserves the intended API contract: max_report_sizes() returns the maximum payload size across all report IDs (excluding the report ID byte), now with correct aggregation.

Potential issues

No issues found.

#	Severity	File	Description	Code
—	—	—	No issues found.	—