v1: data_key_options (CXX-3237, CXX-3238)

[eramongodb]

Resolves CXX-3237 and CXX-3238 for the v1::data_key_options component.

Due to T const& return types in the v_noabi API, v_noabi::options::data_key cannot be refactored in terms of v1::data_key_options. Therefore, only v_noabi <-> v1 conversion functions are defined. Aside from the inlining of most member functions (to reduce the v_noabi ABI footprint), the implementation for v_noabi largely remains unchanged.

Note: view_or_value<T> only exposes a view of its managed object; v_noabi -> v1 is therefore copy-only.

[connorsmacd]

LGTM! Just one comment with minor suggestions.

[connorsmacd]

Suggested change

	for (auto const& name : _impl._key_alt_names) {
	names.push_back(const_cast<char*>(name.c_str())); // For copy only.
	}
	std::transform(
	_impl._key_alt_names.begin(), _impl._key_alt_names.end(), std::back_inserter(names), [](auto const& name) {
	return const_cast<char*>(name.c_str()); // For copy only
	});

I tend to advocate for algorithm use, but with a single-statement loop like this, I chalk it up to preference.

Also, I am not sure what your intentions are with clang-tidy, but you may want to preemptively add a // NOLINT(cppcoreguidelines-pro-type-const-cast).

[eramongodb]

I tend to advocate for algorithm use, but with a single-statement loop like this, I chalk it up to preference.

Accepted the suggestion, with the caveat that generic lambdas are C++14 and newer. I've personally preferred to avoid <algorithm> unless dealing with generic code over iterators (e.g. no direct access to the parent container or knowledge of its type, if any), but there is an argument to be made for opting-into the opportunity for stdlib-internal optimizations based on iterator and element type (likely not applicable here, but nevertheless) and/or ease of extending behavior (e.g. C++17 parallel execution: this could be par_unseq). But maybe this too is YAGNI.

I am not sure what your intentions are with clang-tidy

I'm hesitating to insert any clang-tidy-related comments until we've decided on the set of diagnostics to care about in #1496. Feel free to add yourself as a reviewer if interested in providing input.