Implement the extension GL_AMD_gpu_shader_half_float_fetch

Fix issues of SPIRV headers

- Support new opaque types: f16sampler*, f16image*, f16subpassInput*.
- Add new built-in GLSL texture/image functions.

These are conditional now on which target environment is requested.

Note the tests do not yet generate the correct code regarding what extensions
they declare.

Fix typo in help for -V<bad-number>

EShLangFragment falls through to EShLangCompute, only #ifndef AMD_EXTENSIONS.

EShLangFragment falls through to EShLangCompute, only #ifndef AMD_EXTENSIONS.
This is probably a bug.

This factored computeTypeLocationSize() out of needing the TIntermediate contents,
and uses it to show how to know how many locations an object needs.
However, it still does not do cross stage, or mixed location/no-location
analysis.

Constants were generally cached by type opcode, but all structures share the
same type opcode (OpTypeStruct), so they need to be cached by type id.

I'm not yet clear if these are the right checks, but it puts
the right infrastructure and example in place.

Note: semantic checking against versions/extensions should happen earlier
than SPIR-V generation, so SPIR-V generation need only assume it is okay
to do what was asked.

This means the previously emitted code was incorrect.

Update known-good

See merge request GLSL/glslang!35

GLSL/SPV: Fix #1196: Require resources to have layout(binding=X).

HLSL: Fix #1249: Always execute both sides of ternary "?:".

This is semantically required by HLSL, and frequently results in using
OpSelect instead of control flow.

HLSL: Add conversions for image ops during SPV construction

HLSL allows image and texture types to be templatized on sub-vec4 types,
or even structures.  This was mostly handled already during creation of
sampling operations.  However, for operator[] which can generate image
loads, this wasn't happening.

It also isn't very easy to do at that point in time, because operator[]
does not know where the results it produces will end up.  They may be
an lvalue or an rvalue, and there's a post-process to convert loads to
stores.  They may end up in atomic ops.

To bypass that difficulty, GlslangToSpv now looks for this case and
adds the appropriate conversion.  LIMITATION: this only works for
cases for which a simple conversion opcode suffices.  That is to say,
it will not work if the type is templatized on a struct.

Fix an issue of interpolateAtVertexAMD().

Add --site gitlab to update_glslang_sources command in README

See merge request GLSL/glslang!34

An error message should be reported if the interpolant is not an input.

Update known_good_khr.json to vulkan-1.1-rc3

See merge request GLSL/glslang!33