Add powerpc64 and mips32 architecture support (#170)

* Add powerpc64 and mips32 architecture support

Add startup code for powerpc64le (ELFv2 ABI) and mipsel (O32 ABI) targets,
enabling no_std programs to run on these architectures.

PowerPC64 (powerpc64.rs):
- _start sets up r2 (TOC pointer) using r12 before calling entry
- trap() uses the `trap` instruction
- Full thread support scaffolding (clone, thread pointer, TLS)
- Relocation support for PIC binaries

MIPS32 (mips32.rs):
- __start entry point (MIPS convention, not _start)
- 8-byte stack alignment with 16-byte O32 save area
- trap() uses the `break` instruction
- Thread support using rdhwr for thread pointer
- Syscall convention: $v0=syscall#, $a0-$a3=args, $a3=error flag

Both architectures:
- Implement origin-start feature for program startup
- Support optimize_for_size feature
- Follow existing origin patterns from other architectures

Tested with QEMU user-mode emulation (qemu-ppc64le-static, qemu-mipsel-static).
Note: QEMU ppc64le has a bug with argv setup for non-glibc binaries, but
the startup code itself is correct and works on real hardware.

* Fix rustfmt: remove extra blank line in mips32.rs

That blank line was apparently living rent-free between __start and trap.

* Fix misleading comment about MIPS entry points

We only define __start, not both. The comment was aspirational fiction.

* Simplify mprotect syscall: remove unused r0 output

The syscall return value was just sitting there looking decortaive.
We only need the error flag from $a3.

Note: MIPS cross-compilation is not available on this dev machine,
letting CI handle the actual build verification.

* Use symbolic constant for __NR_set_thread_area

Magic numbers... Now using the proper constant from linux_raw_sys
like civilized folk.

* Fail explicitly for unsupported clone flags on MIPS

O32 ABI requires child_tid (5th arg) on stack, which we don't do.
Check for CLONE_CHILD_CLEARTID/SETTID upfront and return EOPNOTSUPP.

* Trim verbose clone() comment in powerpc64.rs

MIPS got the concise treatment, PPC64 was feeling left out.

* Document which architectures aren't tested in CI

Added a note about arm, powerpc64, and mips32 not being in CI,
along with the reasons why (missing rust-std, crate support gaps,
and a branch that got lost in the shuffle).

Author: walruscraft

README.md

@@ -33,6 +33,20 @@ using nightly Rust, enable the feature "nightly" to let origin use nightly-only
 features, which include proper support for unwinding, better safety checks, and
 better optimizations.
 
+Origin supports multiple architectures: x86-64, x86, aarch64, arm (32-bit),
+riscv64, powerpc64, and mips32. The arm, powerpc64, and mips32 ports are not
+tested in CI, for reasons that range from "Rust doesn't ship binaries for that"
+to "that one crate doesn't support this arch yet":
+
+ - **mips32**: rust-std isn't available for download. CI would need `-Zbuild-std`
+   to build the standard library from source, which is nightly-only.
+
+ - **powerpc64**: On stable, rustix uses the libc backend which is missing the
+   runtime module for signal handling. On nightly, the `unwinding` crate hasn't
+   caught up with ppc64 yet.
+
+ - **arm (32-bit)**: CI setup exists on a branch somewhere, waiting to be merged.
+
 ## Example crates
 
 Origin can also be used on its own, in several different configurations: