The float16 loads and stores are horrendously unoptimized, but this can be addressed as a later pass.

Tests: dEQP-VK.renderpass.suballocation.formats.r16_sfloat.*
Tests: dEQP-VK.renderpass.suballocation.formats.r16g16_sfloat.*
Tests: dEQP-VK.renderpass.suballocation.formats.r16g16b16a16_sfloat.*
Tests: dEQP-VK.pipeline.blend.format.r16_sfloat.*
Tests: dEQP-VK.pipeline.blend.format.r16g16_sfloat.*
Tests: dEQP-VK.pipeline.blend.format.r16g16b16a16_sfloat.*
Bug: b/132434966
Bug: b/132433217
Change-Id: Ifa8feaeecefa1926b1f500e6c9d23e6c242a6844
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/31113Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>