Use the next power of 2 for default_bucket_count

[ArekSredzki]

This allows the compiler to optimize the modulo operation into a bitwise and.

[Bobobalink]

I avoided doing this at first because I was afraid of it causing a large amount of unexpected space overhead, though I agree that it's a good change for runtime reasons.

Do you have any profiling data to show that this is really meaningful? I would be curious to see if the compiler can do other tricks to make modulo faster even for non-powers of two, since this is definitely a compile-time constant.

[alexkaratarakis]

Windows tests are failing:

 D:\a\fixed-containers\fixed-containers\test\fixed_unordered_set_test.cpp(68,32): error C2131: expression did not evaluate to a constant [D:\a\fixed-containers\fixed-containers\build\fixed_unordered_set_test.vcxproj]
      D:\a\fixed-containers\fixed-containers\include\fixed_containers\fixed_robinhood_hashtable.hpp(180,30):
      failure was caused by evaluation exceeding step limit of 1048576 (/constexpr:steps<NUMBER>)