HLSL: support fixed-width types

glslang currently does not support fixed-width types (`float16_t`, `int16_t`, `uint16_t`, `float32_t`, `int32_t`, `uint32_t`). `int64_t` and `uint64_t` are also not supported, as reported in #1346.

The issue can be reproduced with the following code snippet:

```hlsl
#define BUFFER_TYPE uint16_t

[[vk::binding(0)]]
StructuredBuffer in_data;

[[vk::binding(1)]]
RWStructuredBuffer out_data;

[numthreads(32, 1, 1)]
void CSMain()
{
BUFFER_TYPE v = in_data[0];
out_data[0] = v;
}
```

and command:
`glslangValidator -D -e CSMain -S comp --target-env vulkan1.1 --hlsl-enable-16bit-types -H issue.hlsl`

All types compile fine with DXC:
`dxc -E CSMain -T cs_6_6 -spirv -fspv-target-env=vulkan1.1 -enable-16bit-types issue.hlsl`

Start with the issue.hlsl reproduction and run the provided glslangValidator command, then compare its behavior with the shown DXC command. Done means glslang accepts the listed 16-, 32-, and 64-bit fixed-width HLSL types under the relevant Vulkan and 16-bit options.

Written by the indexing model from the issue text.