Inline StepRange construction

[Zentrik]

This can improve performance when iterating over a range with a step as currently steprange_last is not inlined.

using BenchmarkTools

x = rand(Float32, 80)
a = rand(round(Int, length(x) / 2):length(x), 10^6)

function test(x, a)
    c = zero(Float32)

    for j in a
        for i in 1:8:j
            c += x[i]
        end
    end

    return c
end

On 1bf65b9 (similar results on 1.8.5),

julia> @benchmark test($x, $a)
BenchmarkTools.Trial: 380 samples with 1 evaluation.
 Range (min … max):  12.714 ms …  20.363 ms  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     13.091 ms               ┊ GC (median):    0.00%
 Time  (mean ± σ):   13.141 ms ± 470.076 μs  ┊ GC (mean ± σ):  0.00% ± 0.00%

        ▄▄▁  ▁▂█▃▂▅▃▆▇▁▂ ▂                                      
  ▃▁▁▃▃▆████▇██████████████▇▆▄▅▄▃▃▃▃▃▁▃▃▁▁▃▁▁▁▁▁▃▁▁▁▁▁▁▁▃▁▃▃▁▃ ▄
  12.7 ms         Histogram: frequency by time         14.1 ms <

 Memory estimate: 0 bytes, allocs estimate: 0.

With this change

julia> @benchmark test($x, $a)
BenchmarkTools.Trial: 669 samples with 1 evaluation.
 Range (min … max):  7.376 ms …  8.252 ms  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     7.459 ms              ┊ GC (median):    0.00%
 Time  (mean ± σ):   7.465 ms ± 57.103 μs  ┊ GC (mean ± σ):  0.00% ± 0.00%

              ▂▆▂▆▇█▆█▆█▂▇▆▄▇▄▆█▄ ▂▂▂▁▄                       
  ▃▃▂▂▄▄▁▄▆▆████████████████████████████▄▅▄▄▄▄▅▄▄▃▃▄▂▄▂▃▄▂▁▂ ▅
  7.38 ms        Histogram: frequency by time        7.58 ms <

 Memory estimate: 0 bytes, allocs estimate: 0.

While this is a contrived example, if using SIMD.jl you want to loop over 1:8:length(x) and if that's in an inner loop that can't be inlined or if the length is changing this will improve performance.

See https://discourse.julialang.org/t/iterating-over-range-is-slower-than-while-loop/96991 for investigation.

[KristofferC]

Some more conversation in #26770.