That's not accessing memory, so there's no reason it couldn't be (reasonably) efficiently pipelined without decoding to µops.
Not in AArch64 IIRC.
Only the condition code was removed. Otherwise the parent is right and there really is an enormous amount of instructions. Fortunately less than intel, but still a lot.
I don't think you can do variable length shift with an arithmetic instruction in AArch64 either: only immediate shifts.