Building julia on x86_64-linux-musl fails often

[inkydragon]

We may downgrade x86_64-linux-musl to tire 3 supported platforms. JuliaLang/www.julialang.org#2218

Build

The error messages are all memory related:

Collecting and executing precompile statements
└ Collect (Basic: ✓ 1161terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
[11177] signal 6 (-6): Aborted

• https://buildkite.com/julialang/julia-master/builds/43763#01946cec-8bb2-427b-afbf-a22f6d777c0b/845-2694
• https://buildkite.com/julialang/julia-master/builds/43762#01946ceb-e06a-4757-8b0b-5c1350f3af2e/847-2697
• https://buildkite.com/julialang/julia-master/builds/43755#01946bf1-acc8-41ef-a60a-4940154f6d64/845-2694
  LLVM ERROR: out of memory
• https://buildkite.com/julialang/julia-master/builds/43743#01946a19-aadd-476d-88d0-70e52762af0f/845-2694
• https://buildkite.com/julialang/julia-master/builds/43677#01945d12-1557-480e-bec0-3e43422f22bb/845-2821
  LLVM ERROR: out of memory

successful build:

• https://buildkite.com/julialang/julia-master/builds/43741#0194693d-eb0b-4aef-ae8e-424bb7c59fef/844-2814
  successfully precompiled in 409 seconds
• https://buildkite.com/julialang/julia-master/builds/43676#01945cef-a1b6-409f-b2e6-628e26b23958/846-2815
  successfully precompiled in 689 seconds

Maybe we should limit the number of concurrent pre-compilations.

Test

Even if it could be built successfully, it would not pass the test.

	      From worker 7:
2025-01-15 18:13:58 null	      From worker 7:	[7603] signal 11 (1): Segmentation fault
2025-01-15 18:13:58 null	      From worker 7:	in expression starting at none:1
2025-01-15 18:13:58 null	      From worker 7:	pthread_timedjoin_np at /lib/ld-musl-x86_64.so.1 (unknown line)
2025-01-15 18:13:58 null	      From worker 7:	Allocations: 525925194 (Pool: 525923012; Big: 2182); GC: 52

• https://buildkite.com/julialang/julia-master/builds/43741#0194693e-07c7-4df7-9181-a86d19bdce2e/875-1453

LinearAlgebra/triangular                          (3) |  1961.31 |  41.78 |  2.1 |  102854.60 |  1685.92
2025-01-15 18:24:59 null	SparseArrays/sparsevector                        (15) |  1201.61 |  11.77 |  1.0 |   33413.38 |  1249.00
2025-01-15 19:14:39 null	
2025-01-15 19:14:39 null	
2025-01-15 19:14:39 null	Process failed to exit within 5400s, requesting termination and coredump (SIGQUIT) of PID 349.
2025-01-15 19:14:39 null	
2025-01-15 19:14:39 null	[349] signal 3: Quit
2025-01-15 19:14:39 null	in expression starting at none:1

The test times out and terminates, and I wonder if we've hit a deadlock or something similar.