Summary

• build vendored jemalloc with JEMALLOC_SYS_WITH_LG_PAGE=16
• keep jemalloc enabled instead of falling back to the system allocator
• preserve existing jemalloc metrics and profiling support

Context

Users on Linux/aarch64 can hit an immediate abort before any CLI work starts:

<jemalloc>: Unsupported system page size
memory allocation of 4 bytes failed
Aborted (core dumped)

The reported reproducer was spacetime login on Fedora Asahi Remix 42 on Apple Silicon, but the crash happens during allocator init, so it is not specific to login.

This switches from the earlier allocator fallback approach to the narrower fix John requested: build jemalloc with a larger configured page size.

Caveat

Cargo does not support target-specific [env] config, so the repo-level fix uses workspace cargo config and applies to any local build of vendored jemalloc from this repo, not only Linux/aarch64.

Testing

• cargo check -p spacetimedb-cli -p spacetimedb-standalone -p spacetimedb-client-api