Small dispatch fix for gamma

[heltonmc]

Unfortunately we are promoting the argument to Float64 before hitting the gamma calls....

# This PR
julia> @benchmark Bessels.besselj(10, x) setup=(x=rand()*0.0000000001)
BenchmarkTools.Trial: 10000 samples with 996 evaluations.
 Range (min … max):  26.144 ns … 52.565 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     27.193 ns              ┊ GC (median):    0.00%
 Time  (mean ± σ):   27.259 ns ±  0.996 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

                      ▂█▅                                      
  ▂▂▂▂▂▂▂▂▂▁▁▁▁▁▁▁▁▂▂▅███▆▃▄▃▂▁▁▁▁▁▁▁▁▁▁▁▁▁▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▂▁▂ ▂
  26.1 ns         Histogram: frequency by time          29 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

# Master
julia> @benchmark Bessels.besselj(10, x) setup=(x=rand()*0.0000000001)
BenchmarkTools.Trial: 10000 samples with 988 evaluations.
 Range (min … max):  45.090 ns … 98.066 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     46.202 ns              ┊ GC (median):    0.00%
 Time  (mean ± σ):   46.266 ns ±  1.262 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

                    █▂  █▂                                     
  ▂▂▂▂▂▂▂▂▁▁▂▂▂▂▂▂▄███▄███▄▃▃▂▂▂▂▂▂▂▂▂▂▂▂▁▂▁▁▁▁▁▁▁▁▁▁▂▂▂▂▂▂▂▂ ▃
  45.1 ns         Histogram: frequency by time        48.2 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

So this does improve computational time now..... but I'm not for sure we should do much more than this for #58.... hoping our Float32 tests are robust enough to catch if this promotion is even needed.

Edit: Just fixed the dispatch calls to actually hit the gamma integer branch and fixed the proper return value at x = 0. This does nothing to improve #58 as that promotion was needed for accuracy.

[heltonmc]

Ya that is not going to work we need that promotion for single precision. The result for gamma at zero returns NaN instead of Inf for 0 and -Inf for -0.0. Just need a function to return the sign of zero.

[codecov]

Codecov Report

Base: 94.53% // Head: 94.80% // Increases project coverage by +0.27% 🎉

Coverage data is based on head (5ae131e) compared to base (325cb5b).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #63      +/-   ##
==========================================
+ Coverage   94.53%   94.80%   +0.27%     
==========================================
  Files          18       18              
  Lines        1957     1964       +7     
==========================================
+ Hits         1850     1862      +12     
+ Misses        107      102       -5     

Impacted Files	Coverage Δ
src/gamma.jl	86.11% <100.00%> (+13.88%)	⬆️
src/besseli.jl	98.10% <0.00%> (+0.01%)	⬆️
src/besselk.jl	97.28% <0.00%> (+0.02%)	⬆️
src/besselj.jl	93.08% <0.00%> (+0.07%)	⬆️
src/bessely.jl	91.66% <0.00%> (+0.07%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.