ENH: Allow partial table schema in to_gbq() table_schema (#218)

[JohnPaton]

This PR:

• Adds a function to pandas_gbq.schema, update_schema which takes and old schema and a new one, updating any existing fields and adding any new ones
• Generates the default schema using schema.generate_bq_schema for every DataFrame passed to to_gbq, and if a table_schema is passed, this schema is used to update the generated schema. This way custom provided fields in table_schema are added, and missing fields fall back to the generated defaults.

This is the functionality requested in #218

[max-sixty]

This looks great! I think enabling schema 'overrides' is a good feature.

I left one question about the code, but the tests look great, so assuming they pass 👍

[max-sixty]

Do we need this function? Should we import from schema directly? Or there's a circular import?

[JohnPaton]

I was just following the pattern used for the only other schema function, which is imported the same way in gbq._generate_bq_schema. As far as I'm concerned it's fine to get rid of it. Let me know and I'll make the change.

[max-sixty]

I'll defer to Chesterton's fence; we can clean up later if @tswast knows

[tswast]

_generate_bq_schema is only in gbq.py for backwards compatibility. We should use the update_schema method from the schema module directly.

I've sent #259 to clean this up (and also improve the docs for this feature).

[max-sixty]

Great!

Any thoughts @tswast ?

[max-sixty]

@JohnPaton Want to add a Whatsnew? Feel free to add you GH handle

[JohnPaton]

@max-sixty Done! I didn't see any other handles in the Changelog so if that's not what you meant just let me know 😆

[max-sixty]

Many thanks @JohnPaton !

[tswast]

Thanks a bunch for the contribution!