Fix relocs errors on riscv64

[am11]

This ignores the emitter assertion in AOT mode seen here #106223 (comment). Also, missed JALR encoding/decoding in the previous PR. With these changes, the ilc is compiling all the methods and creating object file without errors.

[dotnet-policy-service]

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

[am11]

@dotnet/samsung with these patches,
-r linux-riscv64 -p:PublishAot=true -p:SysRoot=/crossrootfs/riscv64 -p:LinkerFlavor=lld

succeeds. Execution throws SIGILL.

With the default linker (LinkerFlavor=bfd), it doesn't publish. bfd throws:

/usr/bin/riscv64-linux-gnu-ld.bfd: tls_CurrentThread: TLS definition in /runtime/artifacts/bin/coreclr/linux.riscv64.Checked/aotsdk/libRuntime.WorkstationGC.a(threadstore.cpp.o) section .tbss mismatches non-TLS reference in /runtime/artifacts/bin/coreclr/linux.riscv64.Checked/aotsdk/libRuntime.WorkstationGC.a(AllocFast.S.o)

Overall it's a progress. We can take this in and continue hackathon in the tracking issue. PTAL.

[clamp03]

@am11 I removed IMAGE_REL_RISCV64_JALR in #97368 because I think its behavior of IMAGE_REL_RISCV64_JALR and IMAGE_REL_RISCV64_PC looks identical in RISCV. Do you have any reason to add IMAGE_REL_RISCV64_JALR?

[am11]

Thanks. They are handled the same way in ILC as well, only distinguishing themselves at the JIT interface level. I can remove it if the separate identity is not desired.