TensorAlgebra cleanup and optimization

[miguelmaso]

In PR #1 some operations were added in tensor algebra. In this PR, duplicated functions have been removed and optimized.

Additionally, some tests are stronger. For instance, the previous tests were unable to track some operations, given that norm(⊗₁₂³⁴) == norm(⊗₁₃²⁴) == norm(⊗₁₄²³) or norm(⊗₁²³) == norm(⊗₁₂³) == norm(⊗₁₃²). The current tests take advantage of the integer multiplications, which are easy to track component by component.

On the performance side, the substitution of component by component multiplications by generated functions, allow to speedup operations about x30. In the case of contractions in R² space:

BenchmarkTools.Trial: 10000 samples with 158 evaluations per sample.
 Range (min … max):  659.494 ns …  29.168 μs  ┊ GC (min … max): 0.00% … 95.51%
 Time  (median):     735.443 ns               ┊ GC (median):    0.00%
 Time  (mean ± σ):   826.036 ns ± 713.565 ns  ┊ GC (mean ± σ):  8.20% ±  9.00%

  █▇▅▃                                                          ▁
  ██████▅▅▄▁▁▃▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▃▃▃▅ █
  659 ns        Histogram: log(frequency) by time       6.23 μs <

 Memory estimate: 736 bytes, allocs estimate: 7.

BenchmarkTools.Trial: 10000 samples with 998 evaluations per sample.
 Range (min … max):  15.932 ns …  5.511 μs  ┊ GC (min … max):  0.00% … 98.90%
 Time  (median):     19.238 ns              ┊ GC (median):     0.00%
 Time  (mean ± σ):   29.613 ns ± 98.910 ns  ┊ GC (mean ± σ):  14.40% ±  7.46%

  ▄██▇▆▄▂▁                 ▁▂▂▂▂▂▂▂▁▂▂▂▂▁                     ▂
  ███████████▇▆▅▄▄▄▁▅▁▁▁▃▄▆███████████████▇▇▆▆▆▆▅▆▇▆▇▅▆▅▄▅▅▅▄ █
  15.9 ns      Histogram: log(frequency) by time      90.5 ns <

 Memory estimate: 144 bytes, allocs estimate: 1.

The corresponding changes have been applied to the viscous model.

[jmartfrut]

very nice contribution!!

[miguelmaso]

It is passing the tests and I've run some simulations.

I'll wait to merge until you confirm the changes are not affecting your simulations.