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

[eramongodb]

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

The test/private/write_concern.cpp test component is removed: it was only asserting that specific mongoc functions were being invoked by specific mongocxx functions. This behavior is broken by the v1 API + v_noabi refactor, which now uses the *_int64 variants of mongoc_write_concern_set_wtimeout() accessors and avoids calling mongoc_write_concern_set_wmajority() in favor of *_set_w() + *_set_wtimeout() individually. For backward compatibility, .to_document() avoids appending the "wtimeout" field as a $numberLong unless it is strictly necessary, otherwise wc.timeout(0).to_document().view() == scoped_bson{R"({"wtimeout": 0})"}.view() evaluates to false (int32 vs. int64).

[connorsmacd]

LGTM!

Just a few minor comments that reflect my distaste for implicit casts and some stylistic stuff.

[connorsmacd]

Suggested change

	/* explicit(false) */ write_concern(void* impl);
	explicit write_concern(void* impl);

I don't love the idea of an implicit conversion from void*, though one could argue this is acceptable since it's private.

[connorsmacd]

As far as I can tell, MONGOC_WRITE_CONCERN_W_ACKNOWLEDGED is not a real constant in the C Driver, right? If so, I find the use of SCREAMING_SNAKE_CASE here to be a tad confusing. It looks the comment is referring to a constant that does exist, which raises the question as to why the constant was not actually used (if that makes any sense). Obviously, I come from the perspective of someone who is still not terribly familiar with the C Driver.

[connorsmacd]

Suggested change

	return {rc};
	return write_concern(static_cast<void*>(rc));

Suggest using an explicit cast to make it clear which constructor is being called. This comment also assumes you've applied the the suggestion from my other comment. If not, I would still prefer to see {static_cast<void*>(rc)}.

[connorsmacd]

Suggested change

	template struct check<lv1 ::k_default, lv_noabi::k_default>;
	template struct check<lv1 ::k_majority, lv_noabi::k_majority>;
	template struct check<lv1 ::k_tag, lv_noabi::k_tag>;
	template struct check<lv1 ::k_unacknowledged, lv_noabi::k_unacknowledged>;
	template struct check<lv1 ::k_acknowledged, lv_noabi::k_acknowledged>;
	template struct check<lv1::k_default, lv_noabi::k_default>;
	template struct check<lv1::k_majority, lv_noabi::k_majority>;
	template struct check<lv1::k_tag, lv_noabi::k_tag>;
	template struct check<lv1::k_unacknowledged, lv_noabi::k_unacknowledged>;
	template struct check<lv1::k_acknowledged, lv_noabi::k_acknowledged>;

The formatting here does not look correct since I don't see the space after the colon in other places. Maybe it's a clang-format issue?