feat: port node:zlib to rust

[littledivy]

Following a more simpler approach than #18176.

TODO:

• op_zlib_new
• op_zlib_close
• op_zlib_write (and async)
• op_zlib_init
• op_zlib_params
• op_zlib_reset

[littledivy]

# zlib port
        0.02 real         0.01 user         0.00 sys
            33685504  maximum resident set size
                2275  page reclaims
                  37  involuntary context switches
           183640103  instructions retired
            57041452  cycles elapsed
            14042560  peak memory footprint

# main
        0.04 real         0.02 user         0.01 sys
            34308096  maximum resident set size
                2311  page reclaims
                  37  involuntary context switches
           183867121  instructions retired
            55944666  cycles elapsed
            14304640  peak memory footprint

Some improvement in memory usage while there is no visible difference in startup time.

[bartlomieju]

This PR:
Snapshot size: 4_769_424, took 43.374917ms

main branch
Snapshot size: 5_039_504, took 39.408541ms

So we got 270kB saved here. That's a huge improvement 👍

EDIT: Looking at flamegraphs I can see about 2 point reduction in time it takes to deserialize the snapshot and create a context (~59% -> ~57%)

[mmastrac]

Looking good so far.. some early comments.

[mmastrac]

I realize the zlib API is not terribly ergonomic, but bifurcating the compress/decompress paths into separate functions may simplify the code. I don't think this necessarily has to happen, but it would be much easier to understand if the code paths where separated in that way.

[littledivy]

~1ms exec time improvement on Linux

Benchmark 1: deno_main run --reload cli/tests/testdata/run/002_hello.ts 
  Time (mean ± σ):      25.7 ms ±   0.8 ms    [User: 15.8 ms, System: 10.2 ms]
  Range (min … max):    24.2 ms …  28.2 ms    112 runs

Benchmark 1: target/release/deno run --reload cli/tests/testdata/run/002_hello.ts 
  Time (mean ± σ):      24.8 ms ±   1.2 ms    [User: 15.6 ms, System: 9.6 ms]
  Range (min … max):    22.9 ms …  33.3 ms    110 runs

[mmastrac]

LGTM with one small nit