Split ingestion pipeline

[pipliggins]

Splits the ingestion pipeline into two parts; fhirflat file creation and validation.

The initial file creation doesn't pack/unpack any pydantic FHIR resource classes, while validation tries to create resources but doesn't then convert them back into fhirflat files. Together, this provides a significant speedup even when running both file creation and validation together to mimic the previous ingest_to_flat behaviour.

Previous speeds on a sample dataset:

Patient took 1.19 seconds to convert 67 rows.
Encounter took 1.31 seconds to convert 67 rows.
Observation took 6.85 seconds to convert 2194 rows.
Condition took 5.99 seconds to convert 461 rows.

New speeds doing ingestion + validation together (no parallel processing implemented):

Patient took 0.78 seconds to convert 67 rows. 
Encounter took 0.68 seconds to convert 67 rows. 
2 resources not created due to validation errors. Errors saved to encounter_errors.csv
Observation took 1.61 seconds to convert 2194 rows. 
Condition took 0.78 seconds to convert 461 rows. 

Fixes #54

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 98.77049% with 3 lines in your changes missing coverage. Please review.

Project coverage is 97.70%. Comparing base (3f3fa96) to head (6ba5617).
Report is 2 commits behind head on main.

Files	Patch %	Lines
fhirflat/util.py	96.29%	2 Missing ⚠️
fhirflat/resources/base.py	97.14%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #61      +/-   ##
==========================================
+ Coverage   97.58%   97.70%   +0.12%     
==========================================
  Files          39       40       +1     
  Lines        1779     1962     +183     
==========================================
+ Hits         1736     1917     +181     
- Misses         43       45       +2     

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

[abhidg]

Looks good, few minor comments!