[WasmFS] Limit max file descriptor number

[sbc100]

This enabled wasmfs.test_unistd_dup to pass

[kripken]

This looks right, and it enables the test to pass, but it makes WasmFS less capable than before with the new limit.

I guess the limit might be fine, as it's higher than Linux's?

[sbc100]

We can always up the limit or make it configurable via getrlimit/setrlimit.

[sbc100]

According to AI: "The default RLIMIT_NOFILE resource limit for userspace processes on Linux is 1024 (soft) and 4096 (hard). This limit specifies the maximum number of file descriptors that a process can open. "