Fix some i128 shift-related bugs in x64 backend.

[cfallin]

This fixes #2672 and #2679, and also fixes an incorrect instruction
emission (test with small immediate) that we had missed earlier.

The shift-related fixes have to do with (i) shifts by 0 bits, as a
special case that must be handled; and (ii) shifts by a 128-bit amount,
which we can handle by just dropping the upper half (we only use 3--7
bits of shift amount).

This adjusts the lowerings appropriately, and also adds run-tests to
ensure that the lowerings actually execute correctly (previously we only
had compile-tests with golden lowerings; I'd like to correct this for
more ops eventually, adding run-tests beyond what the Wasm spec and
frontend covers).

[bjorn3]

[legion/src/internals/insert.rs:54] component_index = 0
[legion/src/internals/insert.rs:54] 1u128 << dbg!(component_index) = 18446744073709551617

It didn't help unfortunately.

[cfallin]

Ah, yes, you were right; I confused tmp2 for tmp3 (a classic mistake!) and failed to actually use any test vectors with differing lower and upper halves. Should be fixed now (and added a few more test cases). Thanks!

[bjorn3]

Thanks! This version works.

[abrown]

I think this looks fine. (As an aside, why aren't we using SSE2's PSLLDQ/PSRLDQ instructions instead of these long sequences? I haven't looked at much of the i128 code but it would seem that moving upper and lower halves to XMMs and back might still be faster for one of these cases?)

[cfallin]

I think this looks fine. (As an aside, why aren't we using SSE2's PSLLDQ/PSRLDQ instructions instead of these long sequences? I haven't looked at much of the i128 code but it would seem that moving upper and lower halves to XMMs and back might still be faster for one of these cases?)

Ah, the simple answer is that I don't know SSE well enough to reach for such instructions -- though it looks like they should work much more efficiently than these sequences! I'll go ahead and merge with your +1 for now so that we have correct results; but we can definitely improve this later. Thanks!