fix: optimize tables/schema operations

[pnadolny13]

Closes #29

• disable altering for now until we optimize that better. This was also done in target-postgres. I opened feat: support altering efficiently #58 to track adding support back in once it efficient.
• caches table columns and schemas within a sink object. If a schema or key properties change for a stream then a new sink object is created so this is safe to cache and does not need to be invalidated.
• I like the prepare_schema method was used with IF NOT EXISTS as a quick way to avoid these weird column casing and reserved word errors. Its slow to let it constantly retry creating so I removed it by fixing the schema_exists logic to not accidentally try to recreate it again. This required schema names to be conformed before passing to the prepare_schema method.
• UPDATE: there was an issue with the way schema messages were processed that caused github streams to constantly re-initialize new sinks and drain old ones. For the Squared CI test it was doing this 150+ times for a 1 day set of data. I put in a fix here by overriding get_sink to remove _sdc_ columns before comparing because the existing sink schema has already been post processed at that point and the new incoming schema has not. This should be pushed down to the SDK but for now this worked to get my CI tests down to the original transferwise timing. cc @kgpayne

[pnadolny13]

This improved the Squared CI tests enough to finish https://github.com/meltano/squared/actions/runs/5260307866 but the github streams took 27 and 33 mins whereas they took about 7 mins with the transferwise variant. Theres a lot of schema message like I mentioned in the description so I wonder if the logic that diffs the schema against existing ones is accidentally thinking that every schema message is a new schema so its reinitializing a new sink object and draining the previous one, causing lost of wasted time. I'll need to test this theory out.

[kgpayne]

@pnadolny13 looks great 👍 Would be good to create issues for these caches in the SDK.