Add RISCV32 to CI, and use windows-11-arm runners for aarch64-msvc run

[sayantn]

This PR adds CI support riscv32gc-unknown-linux-gnu

Fixes spurious segfaults in Cirrus CI

Adds support for running the aarch64-pc-windows-msvc runners in the new windows-11-arm runners

[rustbot]

r? @Amanieu

rustbot has assigned @Amanieu.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[sayantn]

There seems to be a few CI failures in aarch64-pc-windows-msvc (only in release mode)

disassembly for stdarch_test_shim_test_arm_mte_create_random_tag_irg: 
	 0: 9ac11000
	 1: ret

thread 'core_arch::aarch64::mte::test::assert_test_arm_mte_create_random_tag_irg' panicked at crates\stdarch-test\src\lib.rs:185:9:
failed to find instruction `irg` in the disassembly
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- core_arch::aarch64::mte::test::assert_test_arm_mte_ptrdiff_subp stdout ----
disassembly for stdarch_test_shim_test_arm_mte_ptrdiff_subp: 
	 0: 9ac10000
	 1: ret

thread 'core_arch::aarch64::mte::test::assert_test_arm_mte_ptrdiff_subp' panicked at crates\stdarch-test\src\lib.rs:185:9:
failed to find instruction `subp` in the disassembly

---- core_arch::aarch64::mte::test::assert_test_arm_mte_set_tag_stg stdout ----
disassembly for stdarch_test_shim_test_arm_mte_set_tag_stg: 
	 0: d9200800
	 1: ret

thread 'core_arch::aarch64::mte::test::assert_test_arm_mte_set_tag_stg' panicked at crates\stdarch-test\src\lib.rs:185:9:
failed to find instruction `stg` in the disassembly

---- core_arch::aarch64::neon::assert_vget_high_f64_dup stdout ----
disassembly for stdarch_test_shim_vget_high_f64_dup: 
	 0: umov x8 v0.d[1]
	 1: fmov d0 x8
	 2: ret

thread 'core_arch::aarch64::neon::assert_vget_high_f64_dup' panicked at crates\stdarch-test\src\lib.rs:185:9:
failed to find instruction `dup` in the disassembly

cc @Amanieu. The vget_high_f64 seems to be a generic move inconsistency again, but idk what is going on with the MTE ones (the tests all pass in dev mode)

[Amanieu]

vget_high_f64 is just a move, it should have its instruction test removed.

For the MTE instructions I think the issue is that the MSVC disassembler (dumpbin) doesn't support those instructions.