pred2bq: Update schema parsing from prediction results.

[cfezequiel]

Related to #78

Update code to enable schema parsing from BulkInferrer prediction results data directly.
This adds a third option to the existing ways to parse the input data schema, which include:

1. Parsing from schema.pbtxt file
2. Parsing from TFTransform output path (generated by the TFX Transform component).

Additional changes:

• Revert input of BigQuery table name to use fully qualified name, i.e. 'project:dataset.table_name', instead of specifying GCP project and BigQuery dataset separately.

• Label ID to name mapping is optional - depends on presence of relevant vocabulary file in Transform output

• Typehint and variable name changes

• [ X ] Tests pass

[github-actions]

Thanks for the PR! 🚀

Instructions: Approve using /lgtm and mark for automatic merge by using /merge.

[cfezequiel]

Hi @hanneshapke, I was wondering if you'll be able to review this PR. Sorry for the long-ish change, but it's a bit of a challenge now to break it up into smaller chunks due to time constraints, and subsequent changes also build on top of this PR.

With that said, an alternative would be for me to submit one large PR containing all the subsequent changes for the component.
Pro: One PR to review; total LOC change may be lower than the sum of subsequent PRs due to merging of overlapping changes.
Con: PR would be larger than the previous ones.

Let me know what you think.

[cfezequiel]

This can be deprecated in favor of #230 , which also contains the changes here.

[hanneshapke]

Closing it by request from @michaelwsherman