Don't use musl if you care about performance

The author benchmarked musl’s allocator on 4-core EC2 instances and found it dramatically slower than glibc, with musl-based builds remaining up to 26% slower even when paired with mimalloc. Additional analysis showed that several core memory routines in musl are inherently inefficient, prompting the removal of musl as a prebuilt option for the performance-critical Bifrost service.