Move sqlite Wasm benchmark to first, worst, average scoring

[danleh]

Drive-by: remove unused polyfills, update sqlite source distribution, rebuild.

This now takes about ~6s to run with 15 iterations. I don't want to go much lower with the iteration count, but we could lower --size in argv further (see benchmark.js) to speed up a single iteration. Currently, one iteration takes 300-400ms, so we are well above the threshold we set for ourselves.

@kmiller68 @eqrion @jandem WDYT?

[danleh]

Rebased. @kmiller68 If you are fine with it, feel free to merge (Ryan gave his 👍 above).

[kmiller68]

Given the iterations are so long I wonder if we should change the workload for this? e.g. to a trace from a real sqlite database. Anyway, we can consider that in the future.

[danleh]

If a single iteration is still too long, we can easily change the size parameter in benchmark.js from 10 to, e.g., 5 or 1. (My assumption is that this upstream benchmark provides more coverage of SQLite features and critical code paths than a single trace, given this benchmark is what the SQLite developers use to optimize the database.)