Proposal: ruff + CLI + cloud engines + TPC-DI + Databricks (5-PR stack)

[tomz]

I've got a set of changes I'd like to contribute as a 5-part stack, each based on main:

1. lint — adopt ruff + pre-commit, enforce in CI
2. cloud engines — Livy / Spark Connect + AST-based multi-part name qualification (fixes a silent catalog-dropping bug; makes fuzzy column remap opt-in)
3. cli — lakebench CLI + profiles (fixes a token_env credential-dropping bug)
4. tpcdi — TPC-DI benchmark port across six engines
5. databricks — Databricks Connect engine

Each is independently green (ruff + tests). Do you prefer a base-chained stack of PRs, or sequential PRs against main (open #1, merge, rebase next)? Happy to do whichever fits your review workflow. I'll also call out two design questions inline (the Databricks constructor re-exec, and the W291/W293 lint ignore).