Fix turbulence_noise_scale for Particle Turbulence

[KdotJPG]

Fixes #77491 / my own oversight from #64606/#77154: turbulence_noise_scale needs to have 1 subtracted from it for parity with 4.0.

Curl evaluation in 4.0 here is
curl_3d(noise_pos + noise_time - diff, ...)

= (pos * turbulence_noise_scale - emission_pos) + noise_time - (pos - emission_pos)
= pos * turbulence_noise_scale - emission_pos + noise_time - pos + emission_pos
= pos * turbulence_noise_scale + noise_time - pos
= pos * (turbulence_noise_scale - 1) + noise_time

The position is being effectively multipled by turbulence_noise_scale - 1, but in 4.1 it's only multiplied by turbulence_noise_scale.

The 4.0 formula in set_turbulence_noise_scale is (pow(p_turbulence_noise_scale, 0.25) * 5.6234 / 10.0) * 4.0 - 3.0. This PR changes it to (pow(p_turbulence_noise_scale, 0.25) * 5.6234 / 10.0) * 4.0 - 4.0 to compensate, then simplifies the successive multiplications and divisions to (pow(p_turbulence_noise_scale, 0.25) * 2.24936) - 4.0.

Curve: https://www.desmos.com/calculator/xyzbgqaggw (correctly matches slider range of 0 to 10 again)

[akien-mga]

Thanks for working on fixing that regression! I haven't run the calculations myself but the explanations make sense, so I'm assuming they're correct :)

One suggestion: Should we make these magic numbers more explicit with a comment or a link to a reference where they can be derived from easily? Currently that scaling factor looks pretty arbitrary. (I'm not familiar with that code though so I might be expected here.)

Another workflow related note: We prefer commit titles not to include a direct link to a GitHub URL, but only a plain English description of the bug being fixed. The link to the GH issue should go to the body of the commit message, using one of GitHub's closing keywords (Fixes #77491). E.g.:

Fix particles `turbulence_noise_scale` regression

Fixes #77491.

[KdotJPG]

One suggestion: Should we make these magic numbers more explicit with a comment or a link to a reference where they can be derived from easily? Currently that scaling factor looks pretty arbitrary. (I'm not familiar with that code though so I might be expected here.)

Good idea. We could do something like this:

void ParticleProcessMaterial::set_turbulence_noise_scale(float p_turbulence_noise_scale) {
	turbulence_noise_scale = p_turbulence_noise_scale;
	const float noise_frequency_when_slider_is_zero = 4.0;
	const float max_slider_value = 10.0;
	const float curve_exponent = 0.25;
	const float curve_rescale = noise_frequency_when_slider_is_zero / pow(max_slider_value, curve_exponent);
	float shader_turbulence_noise_scale = pow(p_turbulence_noise_scale, curve_exponent) * curve_rescale - noise_frequency_when_slider_is_zero;
	RenderingServer::get_singleton()->material_set_param(_get_material(), shader_names->turbulence_noise_scale, shader_turbulence_noise_scale);
}

[KdotJPG]

Pushed the above and adjusted commit message.

[YuriSizov]

Thanks a lot!