[Feature] Improve ZipFS.readSync performance

[markdingram]

• I'd be willing to implement this feature

Describe the user story

@aws-cdk/aws-s3-deployment has a 10mb Zip file inside it that hashes on each run with

https://github.com/aws/aws-cdk/blob/a5c06a3a27b39a5315d0cfd0d34b3c1b25cfc464/packages/%40aws-cdk/core/lib/fs/fingerprint.ts

Currently gets stuck due to #1409, after that the it looks like performance is slow due to ZipFS.readSync calling ZipFS.readFileSync afresh on each read into the buffer.

There's a test repo to reproduce - https://github.com/markdingram/yarn-zip

$ yarn run slow
starting
read: 281.731ms 8192 8192
read: 480.655ms 8192 16384
read: 663.234ms 8192 24576
read: 845.223ms 8192 32768
read: 1027.200ms 8192 40960
read: 1210.914ms 8192 49152
read: 1391.626ms 8192 57344
read: 1590.551ms 8192 65536
read: 1764.586ms 8192 73728
read: 1944.802ms 8192 81920
read: 2139.621ms 8192 90112
read: 2333.735ms 8192 98304
read: 2525.436ms 8192 106496
....

Ejecting is a workaround.

Describe the solution you'd like

Cache the file contents in the ZipFS.fds Map on first read to serve subsequent requests.

Describe the drawbacks of your solution

Memory pressure if Zip Files aren't closed after use by existing consumers? Or many files are opened concurrently.

Describe alternatives you've considered

Store an open native File handle in the ZipFS.fds map on first read. Seems more complicated than the proposed approach?

Additional context

Add any other context or screenshots about the feature request here.

Repro - https://github.com/markdingram/yarn-zip

[merceyz]

Closing as fixed since #1743 started caching the read results