include/float8.h:NextPowerOfTwo; infinite values

[SwuduSusuwu]

• +ml_dtypes/include/float8.h:MostSignificantBit; constexpr version of clz opcode.
• *ml_dtypes/include/float8.h:NextPowerOfTwo; replace long list of hardcoded values with 2 << MostSignificantBit<Size>, which extends NextPowerOfTwo (was limited to integer sizes, now is less code plus is general use).

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[SwuduSusuwu]

The original NextPowerOfTwo function (before this pull request) just allows sizes from: 0 through sizeof(int64_t), which can break extended long long and intmax_t types (on architectures where those have more than 8 bytes, which standard C++ does allow).
Now allows sizes from: 0 through LLONG_MAX (so that NextPowerOfTwo has more possible future uses; it is common practice to allow the most possible future uses, as long as to do so does not introduce excessive lines of code (but this pull request reduces the total lines of code)).

[SwuduSusuwu]

Now, NextPowerOfTwo<Size>::value just equals the input value if the input value is a power of two. If the input value is not a power of 2, ::value equals the next power of 2.
Test shows that for all supported input values of the original function, that this new version of the function gives similar output values:

int main() {
  for(int input = 1; 9 > input; ++input) {
    std::cout << "NextPowerOfTwo<" << std::to_string(input) << "> == " << NextPowerOfTwo<1>::value << std::endl;
  }
  return 0;
}

before and after, which now shows similar values for the original versus new.