Remove hardcoded Float64s in AkimaInterpolation

[ChrisRackauckas]

Just doing this from my phone to see if that's it. Missing tests and such, if someone can help that would be great.

[DaniGlez]

Works on the issue example, indeed.

[DaniGlez]

Here's a test:

    # Test consistency between eltype(u) and type of the output
    u = Float32[-0.676367f0, 0.8449812f0, 1.2366607f0, -0.13347931f0, 1.9928657f0,
        -0.63596356f0, 0.76009744f0, -0.30632544f0, 0.34649512f0, -0.3846099f0]
    t = 0.1f0:0.1f0:1.0f0
    for extrapolation_flag in instances(ExtrapolationType.T)
        (extrapolation_flag == ExtrapolationType.None) && continue
        aki = AkimaInterpolation(u, t; extrapolation = extrapolation_flag)
        @test eltype(aki.([-2.0f0, 0.5f0, 3.0f0])) == Float32
    end