perf: spawn jsr futures on executor

[littledivy]

Adds an abstract trait for executing futures onto an async runtime.

A default Tokio spawn executor is included behing the tokio_executor feature flag.

Closes #394

[littledivy]

rustwasm/wasm-bindgen#2562

[littledivy]

I can't test the perf in Deno because it is a few deno_graph versions behind it seems. Are there any other benchmarks I can run? @dsherret

[dsherret]

@littledivy it's up-to-date. Pull the latest main.

[dsherret]

Thanks so much for looking into this!

[dsherret]

Cool, I did not know it was possible to do this for traits.

[littledivy]

2.5x faster on incremental and half peak memory usage:

divy@divy:~/gh/denoland/jsr_benchmarks$ poop 'deno cache --no-check --no-config /home/divy/gh/denoland/jsr_benchmarks/dax_std/mod_jsr.ts' '/home/divy/gh/littledivy/deno/target/release/deno cache --no-check --no-config /home/divy/gh/denoland/jsr_benchmarks/dax_std/mod_jsr.ts'
Benchmark 1 (20 runs): deno cache --no-check --no-config /home/divy/gh/denoland/jsr_benchmarks/dax_std/mod_jsr.ts
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           254ms ±  333ms     177ms … 1.67s           1 ( 5%)        0%
  peak_rss           62.9MB ± 1.38MB    62.1MB … 68.6MB          1 ( 5%)        0%
  cpu_cycles          652M  ±  169M      606M  … 1.37G           1 ( 5%)        0%
  instructions       1.01G  ±  239M      961M  … 2.03G           1 ( 5%)        0%
  cache_references   45.0M  ± 11.4M     42.0M  … 93.5M           1 ( 5%)        0%
  cache_misses       9.77M  ± 4.34M     8.61M  … 28.2M           1 ( 5%)        0%
  branch_misses      3.44M  ± 1.35M     3.09M  … 9.18M           1 ( 5%)        0%
Benchmark 2 (50 runs): /home/divy/gh/littledivy/deno/target/release/deno cache --no-check --no-config /home/divy/gh/denoland/jsr_benchmarks/dax_std/mod_jsr.ts
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           100ms ±  217ms    68.1ms … 1.61s           2 ( 4%)        ⚡- 60.5% ± 53.2%
  peak_rss           39.9MB ± 4.06MB    39.0MB … 68.0MB          1 ( 2%)        ⚡- 36.5% ±  3.0%
  cpu_cycles          200M  ±  136M      177M  … 1.14G           1 ( 2%)        ⚡- 69.4% ± 11.8%
  instructions        310M  ±  208M      281M  … 1.75G           1 ( 2%)        ⚡- 69.4% ± 11.3%
  cache_references   11.9M  ± 9.13M     10.5M  … 75.2M           2 ( 4%)        ⚡- 73.5% ± 11.6%
  cache_misses       3.77M  ± 2.85M     3.25M  … 23.5M           1 ( 2%)        ⚡- 61.4% ± 18.1%
  branch_misses      1.02M  ± 1.05M      844K  … 8.27M           1 ( 2%)        ⚡- 70.4% ± 17.5%

5+ seconds faster with no cache (network is involved):

time /home/divy/gh/littledivy/deno/target/release/deno cache --no-check --no-config -r /home/divy/gh/denoland/jsr_benchmarks/dax_std/mod_jsr.ts

real	0m15.635s
user	0m0.067s
sys	0m0.034s

time deno cache --no-check --no-config -r /home/divy/gh/denoland/jsr_benchmarks/dax_std/mod_jsr.ts

real	0m20.879s
user	0m0.301s
sys	0m0.188s

[dsherret]

LGTM!