Deprecate and remove export to parquet feature

[lmsurpre]

IBM FHIR Server currently has an "export to parquet" feature that is disabled by default.
The feature has seen limited usage and the currently implementation brings in much of Apache Spark which leads oto the following issues when enabled:

1. greatly increases the size of the ibm-fhir-server image
2. increases the attack surface area (e.g. the recent log4j curfuffle)
3. introduces jar hell leading to issues like Odd logging behavior when parquet export is enabled #3070

Unless we can come up with a much better implementation (no small feat), I think we should remove this feature and replace it with documentation that clearly shows how to convert the exported NDJSON to Parquet using spark (maybe a blog post?)

[lmsurpre]

I think we should remove this feature and replace it with documentation that clearly shows how to convert the exported NDJSON to Parquet using spark (maybe a blog post?)

this piece of it is not done yet and I think we should do it either for this one or in a new related task

[lmsurpre]

I split the doc task into its own issue. The spark and stocator depencies have been removed and references to "export to parquet" are removed from the documentation