Align gait aggregation metadata with pulse/tremor outputs and add gait-volume fields

[Copilot]

Gait aggregation output was structurally inconsistent with pulse rate and tremor metadata, and it did not expose key gait-volume context. This update aligns the gait JSON shape and adds explicit metadata for gait amount, filtered gait amount, and arm-swing counts.

• Metadata shape alignment for gait aggregation

  • Adds top-level aggregations["gait"]["metadata"] alongside existing filtered / unfiltered measure blocks.
  • Keeps existing gait digital-measure outputs intact.

• New gait metadata fields (amount + counts)

  • Adds:
    • nr_gait_segments
    • nr_filtered_gait_segments
    • nr_arm_swings
    • nr_filtered_arm_swings
    • gait_duration_s
    • filtered_gait_duration_s
    • hours_of_gait_data
    • hours_of_filtered_gait_data

• Consistent metadata accumulation across files

  • Normalizes empty gait metadata to use combined.duration_s (instead of legacy all.duration_s) in gait pipeline output.
  • Aggregator now safely accumulates combined.duration_s across multi-file runs for both filtered and unfiltered gait.

• Focused regression coverage

  • Adds a gait integration assertion to verify the new aggregations["gait"]["metadata"] shape and duration↔hours consistency.

gait_agg = results["aggregations"]["gait"]

# Existing outputs preserved
filtered_measures = gait_agg["filtered"]
unfiltered_measures = gait_agg["unfiltered"]

# New aligned metadata block
meta = gait_agg["metadata"]
# e.g. meta["hours_of_gait_data"] == meta["gait_duration_s"] / 3600