Add hashtag and editor analytics APIs

[NirrWorks]

Summary

Adds hashtag and editor analytics API support.

Changes

• Added /api/v1/hashtag-stats

  • returns top hashtag leaderboard
  • includes hashtag trends over time
  • supports date range, hashtag filter, interval, limit, and offset

• Added /api/v1/editor-stats

  • returns editor/tool usage stats such as iD, JOSM, StreetComplete, etc.
  • includes changesets, users, map changes, and rank

• Added PostgreSQL indexes for analytics query performance:

  • changesets.hashtags
  • changesets.editor
  • changeset_stats.changeset_id

• Added performance check script using EXPLAIN ANALYZE

  • supports real database testing
  • supports synthetic temporary data testing with large row counts

• Added API tests for new response shapes and SQL behavior

Performance Check

Tested with:

• local DB with ~47k stats rows
• synthetic temporary PostgreSQL dataset with:
  • 1M changesets
  • 1M changeset_stats rows

Synthetic 1M-row results were around:

• hashtag stats: ~3.0s
• hashtag trends: ~2.9s
• editor stats: ~3.2s

This suggests the current queries are acceptable for bounded windows, but for planet-scale/all-time analytics we may later need cached/materialized aggregate tables.

Testing

uv run --group api pytest tests/test_api.py
uv run --group api ruff check api tests scripts/check_api_query_performance.py
uv run --group api ty check api

[kshitijrajsharma]

No need for this file to be included in PR you can keep it in your working tree but not here ! Explain analyze is right way to do it , you can directly do it in database level as well !

[NirrWorks]

Updated the PR with the fixes from the last discussion:

• Fixed hashtag aggregation so hashtag=tt_event aggregates only #tt_event, instead of also returning co-occurring hashtags from the same changesets.
• Fixed empty tag stats behavior so tags=true returns {} when no tag breakdown exists, while tags=false still returns null intentionally.
• Added tests for both cases.

All checks are passing now.

[NirrWorks]

Update

Resolved the merge conflicts with the latest master and added:

• /api/v1/map with changeset centroid points for map clustering and heatmaps.
• Bounding-box export without requiring PostGIS geometry.
• Compact key aggregation by default, with tag_mode=all for full values.
• Optional editor versions using include_version=true.
• Standard pagination metadata across collection endpoints.
• Updated tests, documentation, and performance checks.

[kshitijrajsharma]

No need to write basemodel for pagination try to use the starlette paginator directly !

[kshitijrajsharma]

create index gist ! use sgist index for bbox !
use btree , specify the type of the index