Fix texture flipping for hextiledimage and hextilednormalmap.

[jreichel-nvidia]

Add new mxp_flip_v parameter to forward the generator option for texture flipping.

[jstone-lucasfilm]

This looks like a great improvement to me, @jreichel-nvidia, and I'd ideally like to merge #3001 before reviewing and merging this complementary work, so that we can leverage the new alignment between specification and codebase that it brings.

[jreichel-nvidia]

@jstone-lucasfilm As suggested I've revisited this after #3001 has been merged.

Before this MR (ebc8cc8), the GLSL/OSL and MDL test renderers produces different results.
With this MR (36f5836), GLSL/OSL and MDL agreed.
With main after #3001 (7d0baee), GLSL/OSL and MDL are different.
Merging 7d0baee onto this branch, GLSL/OSL and MDL are still different.

I thought this MR needed some more changes to accomodate for #3001, but could it be that GLSL/OSL is now wrong?

(The following analysis was with some AI support, my apologies if I got it wrong.)

AIUI hex tiling has two distinct steps:

1. From the material’s UV coordinates, determine tile IDs, random rotations/scales/offsets, and blend weights.
2. Sample the image at the resulting coordinates.

Step 1 should produce the same results on every target. Only step 2 needs to account for different image origins.

Let H represent hex tiling and F the V flip. With the current test settings:

 Target               Pattern construction    Coordinates sent to texture lookup
                      (step 1)                (step 2)

 MDL                  H(uv)                   H(uv).coords
 GLSL/OSL (current)   H(F(uv))                H(F(uv)).coords
 GLSL/OSL (correct?)  H(uv)                   F(H(uv).coords)

(MDL uses fileTextureVerticalFlip=false, that's why 'F' it is absent from the 1st row.)

[jreichel-nvidia]

Converted back to draft to make it clear that I do not/no longer intend to merge the PR as it is.