[llvm] [RISCV][MC] Recognise that fcvt.d.s with frm != 0b000 is valid (PR #67555)

[Alex Bradbury via llvm-commits]

asb wrote:

> Do we end up printing the `rne` with no-aliases now? The check lines don't check the whole line so they wouldn't fail if an extra operand started printing.

The InstAlias _should_ mean no (and is working as expected for D), but it seems it's not working as expected for dinx. Agree I need to update this with proper test coverage for that.

I wondered about the adding another FRMArg variant with a different defaultFRMArgOps but instead, but it seemed a shame to add new tablegen classes for this. Maybe it's a better solution overall though - I'll have more of a play with it.

https://github.com/llvm/llvm-project/pull/67555