fix: serve cmocean and crameri colormaps in their published orientation

[matthiasschabel]

Twelve colormaps in the catalog are served in an orientation their upstream projects don't publish — nine cmocean maps stored end-for-end reversed, and three crameri maps still carrying Scientific Colour Maps 7 tables that SCM 8 reversed and revised. This PR fixes the data files; no code changes.

cmocean: nine tables are bit-exact reversals of upstream

Comparing every cmocean:* base entry at 256 samples against the cmocean package's own loader (cmocean 3.0.3, which renders its published gallery), max abs channel difference in 8-bit units:

map	as-is Δ	flipped Δ
algae	212.9	0.0
amp	227.9	0.0
deep	227.4	0.0
dense	226.5	0.0
matter	222.2	0.0
rain	210.8	0.0
speed	231.8	0.0
tempo	233.7	0.0
turbid	215.3	0.0

Flipped Δ = 0 means the stored table is exactly cmocean's, read backwards. The rest of the namespace matches forward exactly (thermal, haline, ice, solar, curl, balance, …, and the cropped balance_blue/curl_pink/delta_blue match cmocean's rgb/*-rgb.py source files as-is), so this is per-map, not a convention difference. The fix reverses the row order in the nine data files; every float literal is unchanged.

crameri: three tables are stale SCM 7 — this completes #83

bilbao, grayC and lajolla are bit-exact matches of SCM 7.0.1 (zenodo 5501399, the __source__ they cite). Fabio Crameri reversed all three in SCM 8 and revised their values beyond the flip — after reversal they still differ from 8.0.1 by up to 35/255 per channel, so a plain flip wouldn't be any published version. #83 updated the namespace to v8's new maps (naviaW here is 8.0.1-exact) but didn't refresh the tables v8 revised.

These three files are regenerated from the SCM 8.0.1 deposit (zenodo 8409685, the same pin naviaW cites), formatted like the #83 additions, __source__ updated.

Reproduce

import numpy as np, cmap, cmocean
grid = np.linspace(0, 1, 256)
ours = np.asarray(cmap.Colormap("cmocean:algae")(grid))[:, :3]
theirs = np.asarray(cmocean.cm.algae(grid))[:, :3]
print(np.abs(ours - theirs).max() * 255)        # ~213 before, ~0 after
print(np.abs(ours - theirs[::-1]).max() * 255)  # ~0 before (bit-exact reversal)

Notes

• This is a visible change for anyone relying on the current orientation of these twelve names; that seemed preferable to permanently diverging from what the upstream projects publish and render in their galleries, but flagging it for the changelog.
• The test suite passes locally (the two okabe_ito matplotlib-parity failures exist on main with matplotlib ≥ 3.10 and are unrelated).
• Happy to split cmocean and crameri into separate PRs if that's easier to review.

🤖 Generated with Claude Code