schema-wasm: optimize for -Oz with LLVM and Binaryen

[aqrln]

This PR explores how much size can be trivially saved in WASM binaries
compared to how we currently build them, using prisma-schema-wasm as
the example. This is not necessarily useful for prisma-schema-wasm
but could be useful for the QE.

• Enable more aggressive size optimization level in LLVM (z instead of
  s) at some runtime performance cost.
• Further optimize the WASM binary for size using wasm-opt from
  Binaryen.

optimization level	size	ratio
-Os (baseline)	2742189	100%
-Oz	2558271	93.29%
-Oz + wasm-opt -Oz	2081699	75.91%

[aqrln]

Marking as draft to make it explicit that it's just a random experiment, not necessarily intended to be merged unless we decide we actually want to do this and have confirmed that performance is acceptable.

[Jolg42]

I think merging this would be great!
The main benefits that I see would be that

• it makes the VS Code extension smaller for all users via making https://www.npmjs.com/package/@prisma/prisma-schema-wasm smaller
• it makes the https://www.npmjs.com/package/prisma Prisma CLI smaller for all users (as it ships the .wasm file).

[codspeed-hq]

CodSpeed Performance Report

Merging #4449 will not alter performance

_{Comparing schema-wasm-optimize-for-oz-with-llvm-and-binaryen (fad616b) with main (e95e739)}

Summary

✅ 11 untouched benchmarks

[aqrln]

Some other numbers from yesterday that I didn't post:

optimization level	size	ratio
-Os	2742189	100%
-Oz	2558271	93.29%
-Os + wasm-opt -Os	2380284	86.80%
-Os + wasm-opt -Oz*	2333130	85.08%
-Oz + wasm-opt -Oz	2081699	75.91%

^{* had to build a bit differently with wasm-bindgen before wasm-opt, otherwise wasm-bindgen was panicking, TODO submit a bug report}

Next steps:

• figure out the performance impact
• maybe start with adding wasm-opt -Os without changing LLVM optimization level at first

[aqrln]

Contributes to https://github.com/prisma/team-orm/issues/584

[janpio]

Removed from DA project based on feedback from Alberto+Serhii.