Vertex2D.raw drops description, state and priority

[Arthur031221]

Vertex2D.raw forwards none of description, state or priority to the
VertexRGB it builds. Volume2D.raw, Volume.raw and Vertex.raw all forward
all three.

import numpy as np, cortex
from cortex.dataset.view2D import Vertex2D

n = cortex.db.get_surf("S1", "flat", merge=True)[0].shape[0]
d1, d2 = np.linspace(0, 1, n), np.linspace(1, 0, n)

raw = Vertex2D(d1, d2, "S1", description="d", state="s", priority=7).raw
print(repr(raw.description), repr(raw.state), raw.priority)

On main at d01dccf this prints '' None 1. The equivalent Volume2D prints
'd' 's' 7.

priority is what orders views inside a Dataset and description is what the
viewer shows, so a figure built from several Vertex2D views comes out labelled
and ordered as though nothing had been set.

Cause

Vertex2D.raw, cortex/dataset/view2D.py:283:

kws = dict(subject=self.dim1.subject)

Volume2D.raw at 199 lists state and description and picks up priority
from **self.attrs.

Fix

Splatting **self.attrs the way Volume2D does is not available here, and
views.py:86 already says why, twice, in this project's own words:

# RGB classes don't accept arbitrary kwargs (e.g. cmap, vmin, vmax),
# so filter to only the parameters they support.
rgb_kwargs = {
    k: v for k, v in kwargs.items() if k in ("description", "state", "priority")
}

VertexRGB takes no **kwargs, so a splat would make Vertex2D(zzz=1).raw
start raising TypeError where it succeeds today. Naming the same three
attributes that block names, the way Vertex.raw does at views.py:518, avoids
that:

kws = dict(subject=self.dim1.subject, state=self.state,
    description=self.description, priority=self.priority)

One regression test, asserting the Volume2D case first as a control so a
failure points at the surface class rather than at the shared colormapping. It
fails on d01dccf and passes with the change. pytest cortex goes from 1 failed
61 passed 55 skipped to 1 failed 62 passed 55 skipped; that failure is
test_warn_non_perceptually_uniform_2D_cmap, which wants Inkscape and fails the
same way before and after.

The alpha guard immediately below is untouched and the alpha arrays are
unchanged either way, which I checked rather than assumed. Vertex2D also does
something questionable with an explicit alpha=, but that runs into the
commented-out block at view2D.py:99 where the interaction between a caller's
alpha, the colormap alpha and NaN vertices is written down and left open, so it
is a separate thing and I would rather measure it properly than bundle it here.

Linux, CPython 3.12.3, numpy 2.5.1, cortex 1.3.3.dev14 editable from d01dccf.
One leg of a matrix covering 3.10 through 3.14, on a machine with no Inkscape,
Playwright or FreeSurfer.