Note: multi-dimension arrays of ConstantBuffer objects will go through uniform flattening.

Fix Android build errors

glslang/MachineIndependent/iomapper.cpp:207:9: error: field 'resolver' will be initialized after field 'stage' [-Werror,-Wreorder]
      : resolver(r)
        ^
glslang/MachineIndependent/iomapper.cpp:263:9: error: field 'resolver' will be initialized after field 'stage' [-Werror,-Wreorder]
      : resolver(r)
        ^
hlsl/hlslParseHelper.cpp:70:5: error: field 'gsStreamOutput' will be initialized after field 'inputPatch' [-Werror,-Wreorder]
    gsStreamOutput(nullptr),
    ^

HLSL: add error for expected comparison sampler in SampleCmp* ops

This adds an error message if a non-comparison sampler is used with
comparison sampling methods.  There's no functional change for correct shaders.

Vector conditions properly convert the true/false expression types to same
width vector as the condition.

Scalar conditions make the true/false expressions convert to each other.

HLSL: fix for byte address buffers in fn parmeters

HLSL: cast non-int types to uint on Load/Store indexes

HLSL: hlsl register class iomapping

Byte address buffers were failing to detect that they were byte address
buffers when used as fn parameters.

Note: this detection is a little awkward, and could be simplified if
it was easy to obtain the declared builtin type for an object.

Adds --hlsl-iomap option to perform IO mapping in HLSL register space.

--shift-cbuffer-binding is now a synonym for --shift-ubo-binding.

The idea way to do this seems to be passing in a dedicated IO resolver, but
that would require more intrusive restructuring, so maybe best for its
own PR.

The TDefaultHlslIoResolver class and the former TDefaultIoResolver class
share quite a bit of mechanism in a common base class.

TODO: tbuffers are landing in the wrong register class, which needs some
investigation.  They're either wrong upstream, or the detection in the
resolver is wrong.

Also, fixed one test file that didn't have its syntax error as the last line.

Some texture and SB operations can take non-integer indexes, which should be
cast to integers before use if they are not already.  This adds makeIntegerIndex()
for the purpose.  Int types are left alone.

(This was done before for operator[], but needs to apply to some other things
too, hence its extraction into common function now)

Longer term, this storage class should be generated based on the mode
of compilation.

HLSL: structuredbuffer counter functionality

This adds TProgram::getUniformBlockCounterIndex(int index), which returns the
index the block of the counter buffer associated with the block of the passed in
index, if any, or -1 if none.

This is WIP, heavy on the IP part.  There's not yet enough to use in real workloads.

Currently present:

* Creation of separate counter buffers for structured buffer types needing them.
* IncrementCounter / DecrementCounter methods
* Postprocess to remove unused counter buffers from linkage
* Associated counter buffers are given @count suffix (invalid as a user identifier)

Not yet present:

* reflection queries to obtain bindings for counter buffers
* Append/Consume buffers
* Ability to use SB references passed as fn parameters

HLSL: Fix #672: Support 1.#INF and -1.#INF syntax.

Null-conversion needs the right sized vectors to kick out with matching types.

HLSL: fix return type for isfinite

The prior decomposition of isfinite was not setting the return type on the
sequence node.  (Sequence was used because there's an internal temporary
to avoid the complex rvalue problem).

Implement the extension SPV_KHR_16bit_storage

HLSL: cast bracket dereference index to int type if not.