[2/🚀] Create feature flag allow-empty-values

[kyle-rader-msft]

This PR address #8, introducing our first feature flag to allow for parsing lines that have a key but no value.

We also enable running the tests with all features enabled as well as the default.

Questions

I'm unsure of what's a better representation for this: Line::Nothing or Pair("non-empty key", "")?

Base PR (#6)

This PR is based on (#6) and I will rebase once that is merged. until then, the diff will be that of both PRs.

[TheDaemoness]

I'm unsure of what's a better representation for this: Line::Nothing or Pair("non-empty key", "")?

Hmm, good question. I like Line::Pair here more on principle as an accurate reflection of the what the line says.
Currently Properties allows empty values to be stored in it, affecting the return value of Properties::len but otherwise only serving to bloat lookup times. One could argue this behavior is a bug 😉. I'll work on fixing it either this weekend or next.

[kyle-rader-msft]

@TheDaemoness , I agree that Pair preserves the most accurate representation, but like you pointed out, is this useful without a value, or is the same interpretation had from pretending it didn't exist at all :).

I've rebased this PR, changing the result to be the Pair with empty value. I also updated the CI to include a second run of the unit tests with the --all-features option enabled to make sure this is covered in CI.

I think your proposed change could be an optimization, but would also be fine without it and letting callers deal with allowing empty values to begin with. The feature flag sort of imples you will get empty values, rather than nonthing ;).

If you wouldn't mind cutting a release after this merges, that would be greatly appreciated!

[TheDaemoness]

Sounds good to me. I'll merge this and, as a temporary measure, alter ConfigParser to discard empty values if the feature flag is enabled, here:

ec4rs/src/parser.rs

Lines 89 to 91 in d469e3e

	Ok(Line::Pair(k, v)) => {
	section.insert(k, v.to_owned());
	}