chore: remove sl_ tables

[mistercrunch]

SUMMARY

I don't know the full history but this collection of models ('sl_.*') were created to support semantic layer, but the work to build on top of it never really started. This PR removes those models and the unfinished things build on top of it.

I dug out this related SIP #17543

It includes:

• sl_dataset_columns
• sl_table_columns
• sl_dataset_tables
• sl_columns
• sl_tables
• sl_dataset_users
• sl_datasets

[john-bodley]

Adding @betodealmeida as a reviewer as he was the original author of said logic.

[betodealmeida]

I don't know the full history but this collection of models ('sl_.*') were created to support semantic layer, but the work to build on top of it never really started. This PR removes those models and the unfinished things build on top of it.

For context, SIP-68 (#14909) proposed to clean up out existing semantic layer, making relationships explicit and differentiating physical datasets from underlying tables.

Since changing the existing models was too big of an effort, the initial code created the new models, and performed a double-write to keep them in sync with the old models. The idea was to update the code to read from the new models, and then drop the double-write and the old models, but this never happened because still the effort was too much.

[mistercrunch]

Could always revert or find inspiration here. But feels good to simplify the erd in the meantime