WasmFS JS API: Implement mmap

[jameshu15869]

This PR copies the changes from #19624 after the LLVM patch (https://reviews.llvm.org/D153466). I decided it might be easier to open a new PR since there were too many merge conflicts to resolve with the original branch back from June.

[kripken]

This code size increase is quite concerning. From the .funcs above it looks like it's due to adding malloc support .

[jameshu15869]

After some changes, it looks like malloc isn't showing up in .funcs, but there's still a slight code size increase. Is this to be expected from the things I've added to library_wasmfs.js?

[kripken]

A small increase with WasmFS+FORCE_FILESYSTEM is expected, but the changes here look odd. For example test/other/test_unoptimized_code_size.js.size doesn't even use WasmFS. Perhaps try --rebaseline after emsdk install tot and emcc --clear-cache to make sure you're building the latest?

[jameshu15869]

I think I've been seeing the code size increases on main as well. I did both emsdk install tot, emsdk activate tot, and emcc --clear-cache. Are you able to reproduce this?

test_unoptimized_code_size (test_other.other) ... Clearing existing test directory
  seen wasm size: 12153 (expected: 12153) (delta: 0), ratio to expected: 0.000000
  seen js size: 60352 (expected: 58664) (delta: 1688), ratio to expected: 0.028774
  seen wasm size: 11600 (expected: 11600) (delta: 0), ratio to expected: 0.000000
  seen js size: 32938 (expected: 31978) (delta: 960), ratio to expected: 0.030021
  seen wasm size: 12153 (expected: 12153) (delta: 0), ratio to expected: 0.000000
  seen js size: 59249 (expected: 57603) (delta: 1646), ratio to expected: 0.028575
ok

----------------------------------------------------------------------
Ran 1 test in 2.106s

OK
(base) PS C:\Users\james\Documents\GitHub\emscripten> git branch
  library-wasmfs-ioctl
  library-wasmfs-mkdev
  library-wasmfs-mmap
  library-wasmfs-mmap-august
  library-wasmfs-mount
  library-wasmfs-mount-acct
  library-wasmfs-readlink-fix
  library-wasmfs-syncfs
* main

[kripken]

I would guess this is the cause of the code size increases. Why is this necessary? I think perhaps we can leave this unchanged but make FORCE_FILESYSTEM in library_wasmfs include this method, as that is the only case it is needed? (since FS.mmap would only be added if FORCE_FILESYSTEM)

[kripken]

Code size issues might have been related to recent landings of LLVM library changes. But that should now be resolved and looks good here.