perf(parser,db): replace lifetime cml_stats refresh with windowed stats

[cchwala]

The background stats timer previously called update_cml_stats() for every known CML every 60 s. That function scans the full history of cml_data, causing 20-30 s of ~100% PostgreSQL CPU usage every minute regardless of how much new data arrived.

This commit replaces that hot path with a cheap windowed variant:

Database (migration 009 / init.sql)

• Add 8 new columns to cml_stats: completeness_percent_6h, total_records_6h, valid_records_6h, mean_rsl_6h, stddev_rsl_6h, completeness_percent_1h, mean_rsl_1h, stddev_rsl_1h
• Add update_cml_stats_windowed(cml_id, user_id) which computes all windowed aggregates in one pass using FILTER clauses. TimescaleDB chunk exclusion limits the scan to the current uncompressed chunk (~6 h of data) irrespective of total dataset size.
• GRANT EXECUTE on the new function to demo_openmrg and demo_orange_cameroun.

Parser (db_writer.py / main.py)

• Add DBWriter.refresh_windowed_stats() mirroring refresh_stats() but calling update_cml_stats_windowed.
• Replace both refresh_stats() call sites in the stats background thread with refresh_windowed_stats().
• Wire _update_stats_for_cmls() into write_rawdata() so lifetime columns (total_records, min_rsl, max_rsl, …) stay current when new data arrives; the stats update and the data insert share one commit so they are atomic.

Webserver (main.py)

• /api/cml-stats: replace the complex LEFT JOIN + live STDDEV recompute with a simple SELECT from cml_stats reading the pre-computed windowed columns. Response now carries completeness_percent_6h as completeness_percent, plus completeness_percent_1h and stddev_last_60min (pre-computed 1 h stddev).
• get_archive_statistics(): replace COUNT(*) FROM cml_data_secure (full hypertable scan) with SUM(total_records) FROM cml_stats (O(users)).

Frontend (realtime.html)

• Update dropdown labels: "Data Completeness" -> "Data Completeness (6h)" and "RSL Std Dev (60min)" -> "RSL Std Dev (1h)".
• Update popup text to match new time-window labels.

Scripts / onboarding (generate_config.py)

• Add GRANT EXECUTE on update_cml_stats_windowed to per-user SQL template so new users get access to both stats functions.

Tests

• test_api_cml_stats.py: rewrite fixture for the new 9-column schema and add assertions for completeness_percent_1h and stddev_last_60min.
• All 63 parser unit tests and 67 webserver unit tests pass.

[codecov]

Codecov Report

❌ Patch coverage is 91.02564% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.42%. Comparing base (ae79bfe) to head (ebc18b7).

Files with missing lines	Patch %	Lines
parser/db_writer.py	76.47%	4 Missing ⚠️
parser/tests/test_main.py	95.00%	2 Missing ⚠️
parser/main.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #51      +/-   ##
==========================================
+ Coverage   84.31%   85.42%   +1.10%     
==========================================
  Files          28       28              
  Lines        2940     3012      +72     
==========================================
+ Hits         2479     2573      +94     
+ Misses        461      439      -22     

Flag	Coverage Δ
mno_simulator	86.12% <ø> (ø)
parser	91.99% <90.90%> (+0.83%)	⬆️
scripts	74.52% <ø> (ø)
webserver	73.78% <100.00%> (+2.99%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.