perf: parse source files in parallel

[dsherret]

See denoland/deno_graph#477

This is much faster on first run, but the same speed on second run (due to the caches).

First run (on mac m2, 8 cores):

% DENO_DIR=$PWD/.cache hyperfine --warmup=3 --prepare="rm -rf .cache/gen .cache/dep_analysis_cache*" "../deno/deno_old run -A main.ts" "../deno/target/release/deno run -A main.ts"
Benchmark 1: ../deno/deno_old run -A main.ts
  Time (mean ± σ):      2.265 s ±  0.013 s    [User: 2.145 s, System: 0.143 s]
  Range (min … max):    2.239 s …  2.283 s    10 runs
 
Benchmark 2: ../deno/target/release/deno run -A main.ts
  Time (mean ± σ):     943.2 ms ±   5.2 ms    [User: 2230.1 ms, System: 146.9 ms]
  Range (min … max):   934.5 ms … 950.2 ms    10 runs
 
Summary
  ../deno/target/release/deno run -A main.ts ran
    2.40 ± 0.02 times faster than ../deno/deno_old run -A main.ts
scratch % cat main.ts
import "https://deno.land/x/ts_morph@22.0.0/mod.ts";
import "https://deno.land/x/ts_morph@21.0.0/mod.ts";
import "https://deno.land/x/ts_morph@20.0.0/mod.ts";
import "https://deno.land/x/ts_morph@19.0.0/mod.ts";

[bartlomieju]

LGTM!