fix: correct 48 drifted prop-table defaults and align 4 variant default mismatches

[Agnik47]

Fixes the two defects reported in #1058. Split into two commits so you can take one and drop the other.

1. fix(docs) — 48 prop-table defaults corrected (15 files, docs only)

propData in src/demo/**/*Demo.jsx had drifted from the components. 48 entries across 16 components documented a value the code does not use. Every corrected value is read from the JS-CSS source, which is the variant the docs page renders.

No runtime behaviour changes in this commit — it only touches the documented values.

The clearest one was TextPressure, whose table was shifted by a row:

Prop	Was documented	Actually
textColor	true (that's stroke's default)	#FFFFFF
strokeColor	#FFFFFF (that's textColor's)	#FF0000
className	#FF0000 (that's strokeColor's)	""

Others of note: ClickSpark had all four of sparkColor / sparkSize / sparkRadius / duration wrong; SideRays had seven props wrong; Lightfall seven; LightRays four.

2. fix(variants) — 3 components where the 4 variants disagreed

CONTRIBUTING rule 5 asks for the variants to stay in sync, but these shipped a different default depending on which variant you installed. In each case 3 of the 4 agree and the documented default agrees with that majority, so I corrected the outlier:

Component	Prop	Change	Was
GlitchText	speed	JS-CSS 1 → 0.5	other 3 variants + docs: 0.5
GlitchText	enableOnHover	JS-CSS true → false	other 3 variants + docs: false
MetaBalls	enableTransparency	JS-CSS true → false	other 3 variants + docs: false
SideRays	falloff	TS-TW 2.0 → 1.6	other 3 variants: 1.6

GlitchText was the user-visible one — install JS-CSS and it glitches continuously; install any other variant and it only glitches on hover.

Preview impact: the GlitchText and SideRays demos pass these props explicitly, so nothing changes on screen. The MetaBalls demo did not pass enableTransparency and relied on the JS-CSS default, so flipping it would have made the preview opaque in light mode — the demo now passes enableTransparency explicitly, keeping the preview exactly as it is today.

Registry items for the three affected variants (GlitchText-JS-CSS, MetaBalls-JS-CSS, SideRays-TS-TW) are rebuilt to match, following the same pattern as the Hyperspeed and ShapeBlur fixes. Each is a one-line change; no other registry file is touched.

Deliberately left alone — your call, not mine

Two splits have no majority and no tiebreaker, so I did not guess:

• FuzzyText fontSize — JS variants use clamp(2rem, 10vw, 10rem), TS variants use clamp(2rem, 8vw, 8rem). A 2-2 split, inconsistent since the component was first added. The docs currently say 8vw.
• LiquidChrome amplitude / frequencyY — JS-CSS uses 0.3 / 3, the other three use 0.5 / 2, and the docs matched neither (0.6 / 1.5). I set the docs to the JS-CSS values since that is what the docs page renders, and left all four sources untouched.

Tell me which value you want in each and I'll follow up.

RippleDistortion's src is also flagged by my scan, but its table deliberately truncates a long Unsplash URL to 'https://images.unsplash.com/...'. Left as-is.

Verification

• npm run lint — 44 problems (33 errors, 11 warnings), identical before and after; all pre-existing.
• npx prettier --check on the touched files — same pass/fail set before and after; no new formatting deviations.
• npx vite build — passes.
• Re-ran the audit script after the changes: 0 remaining mismatches apart from the three cases listed above.

Scope note

My scan could not parse the prop signature of 60 of the 171 components, so this is a lower bound rather than a complete audit. The checker script I offered in #1058 is not included here to keep this PR reviewable — happy to send it separately if you want the class of bug prevented rather than just fixed.

Refs #1058