A few issues with `bit_vector`

[HolyBlackCat]

Pardon me for dumping several issues into a single ticket.

• [x] bit_vector lacks a move constructor.
• [x] The copy assignment should take the parameter by const reference, otherwise this fails:

  const bit_vector<unsigned int> foo();
  bit_vector<unsigned int> v;
  v = foo();

• [x] digits should be static constexpr. There's no reason to store it in each vector instance.
• [x] Move constructor and assignment need to be noexcept, otherwise when placed in a standard container such as std::vector, they'll be copied on resize instead of being moved; see How to enforce move semantics when a vector grows?
• [x] The move assignment leaves the source vector in an inconsistent state, since its vector is now empty but length_ is non-zero. I would zero it. Same applies to the future move constructor. On a side note, the copy constructor an assignment can (and should) be = default;ed.
• [x] Using std::ceil with floating-point division looks odd. It'll give inaccurate results at large sizes, and I expect it to be slower than the integral equivalent: (a + b - 1) / b.
• [x] bit_vector::bit_vector(std::string_view s) does something weird. It interprets the string as bytes, and takes the LSB of each character. It probably should reject everything except '0' and '1', and there should be a separate constructor with a pair of bool iterators.

The rest is purely nitpicking:

• [x] constexpr is missing on a few functions: non-const operator[] and the assignment operators.
• [x] Compiling with -Wextra-semi would give a bunch of warnings on ;s after closing braces.

[bkille]

Wow, thanks for the detailed feedback! All great points I'll address them ASAP

[bkille]

@HolyBlackCat I think I've addressed all of these in my most recent PR #41 (with the defaults being added directly to master). Please re-open if you see anything amiss and again, thank you for reviewing this!