COST-953: GCP processing window fix

[dccurtis]

The GCP downloader _generate_monthly_pseudo_manifest function was generating a scan window that was one day off. Additionally there was an assumption that the partitioning date is the same as the usage date. This did not turn out to be the case and we are now doing the BigQuery query based on usage_start so that it aligns with the summary processor

Testing

1. Set DATE_OVERRIDE to Jan-29 and ingest GCP. Verify data comes in for 1-29 of January
2. Set DATE_OVERRIDE to Jan-30 and ingest GCP. Verify data comes in for 1-30 of January. Verify 1-29 does not change.
3. Set DATE_OVERRIDE to Jan-31 and ingest GCP. Verify data comes in for 1-31 of January. Verify 1-30 does not change.
4. Set DATE_OVERRIDE to Feb 1 and ingest GCP. Verify data comes in for 1 January - 1 Feb. Verify January data does not change.

Test Results
cost_593_ut.txt

[codecov]

Codecov Report

Merging #2622 (ec372d4) into master (b8fbc68) will decrease coverage by 0.0%.
The diff coverage is 100.0%.

@@           Coverage Diff            @@
##           master   #2622     +/-   ##
========================================
- Coverage    94.8%   94.7%   -0.0%     
========================================
  Files         281     281             
  Lines       21527   21529      +2     
  Branches     2445    2445             
========================================
- Hits        20397   20393      -4     
- Misses        679     682      +3     
- Partials      451     454      +3     

[myersCody]

Sweet glad to see the switch to usage_start_time < '{scan_end}' worked!

[myersCody]

Awesome work on this @dccurtis!