Correction to report splitting

[MattWellie]

Fixes

• Closes #421
• Closes #422
• When generating multiple reports for a cohort (splitting by year, or an arbitrary split), I was deriving the output folder to write additional reports to from Path(<main_output_file>).parent
• The use of Pathlib.Path on a gs://... address removes the double //
• cloudpathlib interprets a gs:/... prefix to be a local path, and can't open the requested file path
• Result was that for the cohorts we now split (due to being... massive) we now write the main and latest report, then the report generation script breaks, so we don't get the mini-reports

Second issue - #421

In the Hail analysis of SV data, the process quits if there's no overlap between the PED and VCF

This has caused recurrent failures where a participant has both Exome and Genome results - only the exome data is being retained in the pedigree, so there appears to be no overlap with the exome-only CNV data

Proposed Changes

• Stop using Pathlib.parent to get the directory name
• Instead just get the file name, and remove from the full path to get the directory
• use plain string manipulation to build the path after that

Checklist

• [x] Related Issue created
• [ ] Tests covering new change
• [x] Linting checks pass

[MattWellie]

also took this opportunity to delete more linting exclusions