HLSL requires vec2 tessellation coordinate declarations in some cases
(e.g, isoline topology), where SPIR-V requires the TessCoord qualified
builtin to be a vec3 in all cases.  This alters the IO form of the
variable to be a vec3, which will be copied to the shader's declared
type if needed.  This is not a validation; the shader type must be correct.

Somewhat complex due to recognizing a general scalar, but not
replicating it for each member to avoid side effects.

HLSL: Decompose OpIsFinite to avoid capability restrictions

Improves foundation for adding scalar casts.

Makes handle/make names more sane, better commented, uses more
precise subclass typing, and removes mutual recursion between
converting initializer lists and making constructors.

HLSL: Support InputPatch variables in patch constant functions

OpIsFinite is not available everywhere: decompose isfinite
to !isinf && !isnan.

Previously, patch constant functions only accepted OutputPatch.  This
adds InputPatch support, via a pseudo-builtin variable type, so that
the patch can be tracked clear through from the qualifier.

HLSL: fix GS to work with EP wrapping

The prior implementation of GS did not work with the new EP wrapping architecture.
This fixes it: the Append() method now looks up the actual output rather
than the internal sanitized temporary type, and writes to that.

HLSL: handle PCF input to DS in arbitrary argument position

In the hull shader, the PCF output does not participate in an argument list,
so has no defined ordering.  It is always put at the end of the linkage.  That
means the DS input reading PCF data must be be at the end of the DS linkage
as well, no matter where it may appear in the argument list.  This change
makes sure that happens.

The detection is by looking for arguments that contain tessellation factor
builtins, even as a struct member.  The whole struct is taken as the PCF output
if any members are so qualified.

Refactor TType::contains* methods (nonfunctional)

There were many (~8) different places in TType which all knew how to traverse
the struct/type hierarchy.  There's a need to add another, but I didn't
want to duplicate the traversal code again.  This is a small refactoring
which passes a predicate to a single traverse-and-test method.  That also
shortens all the containsSomething() methods from 9 lines of body to 1.

There are no test differences: it's nonfunctional.

HLSL: set combined flag false for Buffer<>

This reverts commit 1dd65ca3, reversing
changes made to 4960baaf.

HLSL: support per control point patch const fn invocation

HLSL: Fix #804: crash on empty sequence node passed to intrinsic expansions

The SPIR-V generator had assumed tessellation modes such as
primitive type and vertex order would only appear in tess eval
(domain) shaders.  SPIR-V allows either, and HLSL allows and
possibly requires them to be in the hull shader.

This change:

1. Passes them through for either tessellation stage, and,

2. Does not set up defaults in the domain stage for HLSl compilation,
to avoid conflicting definitions.

HLSL: allow length() on scalars

Covers if(cond), while(cond), do-while(cond), for(;cond;), and (cond ? :).
Fixes #778.

Unknown how extensive the semantics need to be yet. Need real
feedback from workloads. This is just done as part of unifying it
with the class/struct namespaces and grammar productions.

HLSL HS outputs a per ctrl point value, and the DS reads an array
of that type.  (It also has a per patch frequency).  The per-ctrl-pt
frequency is arrayed on just one side, as opposed to SPIR-V which
is arrayed on both.  To match semantics, the compiler creates an
array behind the scenes and indexes it by invocation ID, assigning
the HS return value to it.

SPIR-V requires that tessellation factor arrays be size 4 (outer) or 2 (inner).
HLSL allows other sizes such as 3, or even scalars.  This commit converts
between them by forcing the IO types to be the SPIR-V size, and allowing
copies between the internal and IO types to handle these cases.

This PR emulates per control point inputs to patch constant functions.
Without either an extension to look across SIMD lanes or a dedicated
stage, the emulation must use separate invocations of the wrapped
entry point to obtain the per control point values.  This is provided
since shaders are wanting this functionality now, but such an extension
is not yet available.

Entry point arguments qualified as an invocation ID are replaced by the
current control point number when calling the wrapped entry point.  There
is no particular optimization for the case of the entry point not having
such an input but the PCF still accepting ctrl pt frequency data.  It'll
work, but anyway makes no so much sense.

The wrapped entry point must return the per control point data by value.
At this time it is not supported as an output parameter.

HLSL: Fix an issue of frexp().

Don't emit NV-specific interface members if NV extensions are disabled.

The "exp" parameter is floating-point type in HLSL intrinsic while it is
integer type in GLSL built-in function.