Fix workgroup builtins size

NumWorkgroups, WorkgroupId, and WorkgroupSize are uniform globals with
three components. While we store them in a 4-component vector, they must
not be stored in a SIMD::Int.

PointCoord was corrected to a 2-component vector (per lane).

The X, Y, and Z enums were replaced by explicit numeric indexes. Also
the loops for multi-component builtins were replaced with explicit
assignment of each component. The asserts for single-component builtins
have been removed since these are checked by spirv-val.

Bug: b/214583550
Change-Id: Iecc81f1f255978e160712fbaa0f069b8e7065705
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/66510
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

src/Pipeline/ComputeProgram.hpp

diff --git a/src/Pipeline/ComputeProgram.hpp b/src/Pipeline/ComputeProgram.hpp
index 3fc6d97..9af300c 100644
--- a/src/Pipeline/ComputeProgram.hpp
+++ b/src/Pipeline/ComputeProgram.hpp
@@ -72,8 +72,8 @@
 	{
 		vk::DescriptorSet::Bindings descriptorSets;
 		vk::DescriptorSet::DynamicOffsets descriptorDynamicOffsets;
-		uint4 numWorkgroups;               // [x, y, z, 0]
-		uint4 workgroupSize;               // [x, y, z, 0]
+		uint4 numWorkgroups;               // [x, y, z, -]
+		uint4 workgroupSize;               // [x, y, z, -]
 		uint32_t invocationsPerSubgroup;   // SPIR-V: "SubgroupSize"
 		uint32_t subgroupsPerWorkgroup;    // SPIR-V: "NumSubgroups"
 		uint32_t invocationsPerWorkgroup;  // Total number of invocations per workgroup.