Add FastInterpolations as a faster resampling backend

[mgyoo86]

Hi!
I'm the author of FastInterpolations.jl, a new high-performance interpolation package, and I'm looking for more places to use it.
I found that Resampler (and its Pyramid payload) looks like a nice fit; it resamples on every pan/zoom, so faster interpolation helps interactivity.

This PR adds a small, shallow wrapper so Resampler (and its Pyramid payload) can use the FastInterpolations API through the existing method/mode keyword.

In the real per-frame resampling path, FastInterpolations is 2–9× faster than Interpolations.jl and matches the current output to floating-point round-off. So I made FastInterpolations.LinearInterp() the default; anything that already used Interpolations directly keeps working, and reverting the default is a one-line change.

In addition, the curvilinear contourf remap now also uses FastInterpolations — internal only, with identical output.

Benchmarks

Interpolations.jl's Linear() vs. the new default FastInterpolations.jl's LinearInterp(), measured through the real resample_image (the per-frame function):
a (2000x2000) source, either a Float32 field and an RGB{N0f8} photo-like image, downsampled to (500x500) for display.

Payload	Interpolations	FastInterpolations	Speedup
Resampler Float32	1.25 ms	0.20 ms	6.1×
Resampler RGB{N0f8}	2.20 ms	0.98 ms	2.2×
Pyramid Float32	1.64 ms	0.19 ms	8.6×
Pyramid RGB{N0f8}	4.10 ms	0.59 ms	6.9×

The following is a benchmark script for the table above:

using Makie, BenchmarkTools, Colors
import FastInterpolations as FI
using Makie: Interpolations
const N0f8 = Colors.N0f8

N = 2000
f32 = Float32.(sin.((1:N) ./ 40) * cos.((1:N) ./ 50)')                       # outer product
rgb = RGB{N0f8}.(abs.(sin.((1:N) ./ 40)), abs.(cos.((1:N) ./ 50))', 0.4)     # real-photo type
X, Y = (1.0f0, Float32(N)), (1.0f0, Float32(N))
lim = Makie.Rect2f(1, 1, N - 1, N - 1)  
res =  (500, 500) # 2000² → 500² (4× downsample)

for (name, payload) in [
        "Resampler Float32 / Interpolations" => Makie.Resampler(f32; method = Interpolations.Linear()).data,
        "Resampler Float32 / FastInterp"     => Makie.Resampler(f32).data,
        "Resampler RGB / Interpolations"     => Makie.Resampler(rgb; method = Interpolations.Linear()).data,
        "Resampler RGB / FastInterp"         => Makie.Resampler(rgb).data,
        "Pyramid Float32 / Interpolations"   => Makie.Pyramid(f32; mode = Interpolations.Linear()),
        "Pyramid Float32 / FastInterp"       => Makie.Pyramid(f32),
        "Pyramid RGB / Interpolations"       => Makie.Pyramid(rgb; mode = Interpolations.Linear()),
        "Pyramid RGB / FastInterp"           => Makie.Pyramid(rgb),
    ]
    print(rpad(name, 36), ": ")
    @btime Makie.resample_image($X, $Y, $payload, $res, $lim)
end

This PR needs FastInterpolations 0.4.16, pinned in [compat].
Thanks for taking a look, any feedbacks and thoughts are welcome!

[mgyoo86]

For the maintainers' information:
the previous run (1333d98) passed all CI checks except the changelog one. I've now added a commit that only updates CHANGELOG.md, with no source code changes, so a full CI re-run probably isn't necessary.
Thanks!

[ffreyer]

/benchmark 30

Oh, nvm, benchmarks don't work with branches on forks. I wanted to check if this influences using time, just in case

[mgyoo86]

For maintainters' information:
I’ve merged the latest master branch into this PR to resolve the conflict in the CHANGELOG.
(no conflicts in the actual source codes)

@ffreyer
I don’t think this will cause any performance issues, but… yes, it would have been better if the benchmark CI had been able to run on this PR.