Fix quant/dequant table bug

[FioraAeterna]

1<<31 is not 2^31 because 1 is a signed int in C; this bug affected both the
JIT and interpreter quantized store implementations, though I don't know if
any games were actually affected.

[degasus]

Oh, I've thought you're talking about this table: https://github.com/dolphin-emu/dolphin/blob/master/Source/Core/VideoCommon/VertexLoader.cpp#L67

Why do we have this table so often in memory?

imo it's better to always write "1u << xx", else LGTM

[FioraAeterna]

The table's in memory once for the interpreter and once for the JIT, I think.

[FioraAeterna]

Changed the tables to a consist 1ULL on Jasper's request.

[degasus]

@FioraAeterna Nope, all of them: 1ULL << 0, 1ULL << 1, 1ULL << 2, ...

[degasus]

LGTM

But tbh, isn't there a nicer way to add an integer to the exponent but a multiplication? Let's hope SSE didn't miss this kind of bit operations. But of course of of scope of this PR.

[FioraAeterna]

I'm not sure there's a nicer way; adding/subtracting to the exponent directly isn't any fewer uops and might break terribly in cases like 0, denormals, and so on. I don't think SSE has an equivalent of an integer bitshift on a float.