feat: Add compression option ZSTD.

[chalmerlowe]

Based on a PR submitted by EthanSteinberg.

One of BigQuery's neat features is how it supports ZSTD compression for exports.
See: https://cloud.google.com/bigquery/docs/exporting-data#parquet_export_details
This commit simply adds ZSTD to the list of enums allowed for the compression type.

I added a test to confirm that the enum is correctly populated with only the allowed options.

For future me, including this link directly to the list of current export formats and the allowable compression types.

Closing Ethan's PR.

[chalmerlowe]

Before we merge, there are a couple items I wanna investigate.
More to come.

[tswast]

I wonder how useful this test is? Seems an awful lot like a change-detector test to me. I'd be fine adding the constant without adding the test.

Alternatively, maybe there's a system test we could write to make sure this is synced with the bigquery discovery document? But even then, compression isn't a true enum. The allowed values are only listed in the documentation string from what I can tell.

[chalmerlowe]

@tswast

I will remove this test.

In terms of attempting to ensure that this matches the discovery docs... the terms referenced by the docs are present in the description, as you note, so we would need some mechanism to extract them from the discovery doc, which feels somewhat fragile (ie extract all words that are ALL CAPS and deduplicate them). Thoughts?

"JobConfigurationExtract": {
      ...
      "properties": {
        "compression": {
          "description": "Optional. The compression type to use for exported files.
Possible values include DEFLATE, GZIP, NONE, SNAPPY, and ZSTD. The
default value is NONE. Not all compression formats are support for all
file formats. DEFLATE is only supported for Avro. ZSTD is only supported
for Parquet. Not applicable when extracting models.",
          "type": "string"
        },

[tswast]

Agreed. Without a structured representation of the allowed values, it's too fragile.