fix: NoiseEffectController producing zero progress on some platforms

[ufrshubham]

Description

NoiseEffectController was sampling the noise function with a hardcoded y-coordinate of 1:

noise.getNoise2(timer / duration, 1)

PerlinNoise(frequency: F) scales inputs internally by F before sampling, so the effective y becomes 1 * F = F. With any integer frequency (e.g. frequency: 400), y lands on an integer row in the noise grid.

When this happens, yd0 = y - y0 = 0 (fractional distance from the lower lattice row) and consequently ys = 0 (the quintic interpolation weight). The ys.lerp(bottomRow, topRow) call then collapses to just bottomRow, discarding all top-row gradient contributions.

Within the surviving bottom row, gradCoord2D computes xd * g.x + yd * g.y. Since yd0 = 0, the y-gradient term vanishes entirely. Additionally, 2 of the 8 possible gradient vectors — (0,-1) and (0,1) — have g.x = 0, so any corner the hash maps to those contributes nothing. The result is a severely reduced noise amplitude, sometimes collapsing to near-zero depending on the hash outcome for the seed and x-coordinate.

Fixes:

• Changed y from 1 to sqrt2 (≈ 1.41421356…). Since sqrt2 is irrational, no integer frequency can scale it to an integer row, so yd0 ≠ 0, ys ≠ 0, and all 4 corner gradients (both x and y components) always contribute fully.
• Changed x from timer / duration to timer so frequency correctly means oscillations per second regardless of effect duration. Previously, normalizing to 0..1 meant duration was silently controlling oscillation speed.

Checklist

• I have followed the Contributor Guide when preparing my PR.
• I have updated/added tests for ALL new/updated/fixed functionality.
• [NA] I have updated/added relevant documentation in docs and added dartdoc comments with ///.
• [NA] I have updated/added relevant examples in examples or docs.

Breaking Change?

• Yes, this PR is a breaking change.
• No, this PR is not a breaking change.

Related Issues

Closes #3193

[spydon]

Lgtm!