Docs direct users to slow options

This may not need to be changed, but just FYI `LinearInterpolation` is *not* free of extrapolation, it merely defines extrapolation with `Throw()` behavior. And this does not come cheap:
```julia
julia> using Interpolations, BenchmarkTools

julia> xs = 1:0.2:5
1.0:0.2:5.0

julia> f(x) = log(x)
f (generic function with 1 method)

julia> A = [f(x) for x in xs];

julia> itp = LinearInterpolation(xs, A); # what the README directs people to use

julia> itp2 = scale(interpolate(A, BSpline(Linear())), xs); # a lower-level alternative

julia> function interpolate_many(itp, iv, n)
s = zero(itp(minimum(iv)))
xeval = minimum(iv) .+ (maximum(iv) - minimum(iv)).*rand(eltype(iv), n)
@inbounds @simd for i = 1:n
s += itp(xeval[i])
end
return s
end
interpolate_many (generic function with 1 method)

julia> @benchmark interpolate_many(itp, xs, 10000)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
Range (min … max): 51.328 μs … 865.010 μs ┊ GC (min … max): 0.00% … 93.11%
Time (median): 56.020 μs ┊ GC (median): 0.00%
Time (mean ± σ): 63.901 μs ± 36.250 μs ┊ GC (mean ± σ): 2.71% ± 4.93%

▂▇▇█▆▁▁▂▂▂ ▁▂▃▃▃▂▂▂ ▂
████████████▆▆▅▆▇▆█▇▆▅▃▄▃▄▆▅▅▄▂▄▃▅▄▄▅▇█████████▇▆▆▆▅▃▅▅▅▄▅▃▄ █
51.3 μs Histogram: log(frequency) by time 113 μs <

Memory estimate: 156.42 KiB, allocs estimate: 5.

julia> @benchmark interpolate_many(itp2, xs, 10000)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
Range (min … max): 33.864 μs … 910.252 μs ┊ GC (min … max): 0.00% … 93.33%
Time (median): 36.828 μs ┊ GC (median): 0.00%
Time (mean ± σ): 44.396 μs ± 38.660 μs ┊ GC (mean ± σ): 4.35% ± 5.26%

▃▇█▆▂▁ ▂▂▁ ▃▃▂▂▂▃▂▁ ▂
████████████▇▆▅▅▅▅▅▆▇█▇▅▅▅▃▃▃▁▅▅▃▃▁▄▃▅▃▄▄▅▁▄▄█████████▇▆▇▆▆▇ █
33.9 μs Histogram: log(frequency) by time 84.6 μs <

Memory estimate: 156.42 KiB, allocs estimate: 5.
```

Let's also try with integer-indexes:
```julia
julia> itpi = LinearInterpolation(1:length(A), A);

julia> itpi2 = interpolate(A, BSpline(Linear()));

julia> @benchmark interpolate_many(itpi, 1.0:length(A), 10000)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
Range (min … max): 51.644 μs … 934.393 μs ┊ GC (min … max): 0.00% … 92.43%
Time (median): 54.011 μs ┊ GC (median): 0.00%
Time (mean ± σ): 57.281 μs ± 25.354 μs ┊ GC (mean ± σ): 2.35% ± 5.08%

▃▅██▅▁ ▁
██████▇▇▇▇▅▆▅▆▅▄▅▅▅▄▅▅▄▄▄▄▄▃▄▄▃▂▂▃▄▃▄▄▄▄▃▂▅▅▅▅▅▅▆▆▇▇▆▆▇██▇▅▅ █
51.6 μs Histogram: log(frequency) by time 98.2 μs <

Memory estimate: 156.45 KiB, allocs estimate: 6.

julia> @benchmark interpolate_many(itpi2, 1.0:length(A), 10000)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
Range (min … max): 20.348 μs … 557.423 μs ┊ GC (min … max): 0.00% … 90.11%
Time (median): 24.013 μs ┊ GC (median): 0.00%
Time (mean ± σ): 26.495 μs ± 25.515 μs ┊ GC (mean ± σ): 5.28% ± 5.32%

▂▁ ▄██▅▂ ▂
██▇██████▇██▇▆▆▅▆▆▅▄▅▅▅▆▆▅▅▄▅▃▅▄▁▃▁▅▁▁▁▄▁▁▅▄▅▃▅▅▃▁▅▃▅▄▃▆██▇▇ █
20.3 μs Histogram: log(frequency) by time 63.9 μs <

Memory estimate: 156.45 KiB, allocs estimate: 6.
```

The difference is the branch, and SIMD-vectorization. My laptop is AVX2, an AVX512 machine would show even larger differences.