fix(driver-adapters): PlanetScale transactions

[jkomyno]

This PR deprecates #4966.

This PR:

• Contributes to [DA] Planetscale engine tests: interactive_tx #4469
• Contributes to https://github.com/prisma/team-orm/issues/953

Companion TypeScript PR: prisma/prisma#24878.
Please refer to the explanations therein about TransactionContext.

[codspeed-hq]

CodSpeed Performance Report

Merging #4967 will not alter performance

_{Comparing integration/fix-planetscale-transactions (0deadde) with main (f2561ec)}

Summary

✅ 11 untouched benchmarks

[github-actions]

WASM Query Engine file Size

Engine	This PR	Base branch	Diff
Postgres	2.064MiB	2.062MiB	1.988KiB
Postgres (gzip)	824.513KiB	823.313KiB	1.201KiB
Mysql	2.034MiB	2.032MiB	2.117KiB
Mysql (gzip)	811.424KiB	811.001KiB	433.000B
Sqlite	1.925MiB	1.924MiB	1.952KiB
Sqlite (gzip)	769.174KiB	768.301KiB	894.000B

[aqrln]

It's kind of a slippery slope but I really feel like implementing Deref is appropriate in these cases. Let's not do it in this PR to avoid churning unrelated code, but I'd strongly consider doing it later. Others may disagree though.

[aqrln]

Although I was sceptical at fist, after thinking about it a bit more I agree that it's an appropriate abstraction. We still have a kind of typestate pattern (i.e. TransactionContext doesn't have methods like commit/rollback, you need to call start_transaction to get access to them) while the Queryable trait represents another dimension and is orthogonal to those methods.

While ideally I would still prefer to have tighter interfaces or push more logic to the driver adapter without going back and forth between JS and Rust so much, that would be a big refactoring, while this PR is indeed a good solution based on the abstractions we have.