avoid restricting types to real

[GiggleLiu]

See #98

[chriselrod]

You'll need to resolve the ambiguities.

[GiggleLiu]

I just fixed the ambiguity error by seperating the vector and matrix implementations of Complex matmul, does it look good?

[codecov]

Codecov Report

Merging #99 (ef755f1) into master (5ded095) will decrease coverage by 1.59%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #99      +/-   ##
==========================================
- Coverage   87.19%   85.59%   -1.60%     
==========================================
  Files          11       11              
  Lines         742      743       +1     
==========================================
- Hits          647      636      -11     
- Misses         95      107      +12     

Impacted Files	Coverage Δ
src/block_sizes.jl	96.66% <100.00%> (ø)
src/complex_matmul.jl	100.00% <100.00%> (ø)
src/macrokernels.jl	89.55% <100.00%> (ø)
src/matmul.jl	90.65% <100.00%> (-0.35%)	⬇️
src/global_constants.jl	42.00% <0.00%> (-18.00%)	⬇️
src/funcptrs.jl	97.05% <0.00%> (-2.95%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5ded095...ef755f1. Read the comment docs.

[chriselrod]

It may be nicer to rename these complex _matmul methods, and then have the specific _matmul methods you're @evaling here simply forward to those renamed methods, but this is fine.
Thanks.