Fixes out-of-bounds reads/writes and also avoids unaligned accesses.

This is a generic reference implementation. Future optimizations could
use unaligned 4-byte accesses if they're known to be safe (e.g. due to
padding) and efficient. We can also eliminate if we know we're in a
basic block post-dominated by the entry block and the number of
active invocations is a multiple of the SIMD width.

Bug: b/160531165
Change-Id: I892cfd3c7da8d8891cabe80695e5f35c57da73b4
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/46168
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>