Fix parsing of unquoted values with two spaces

[bbc2]

If a value is unquoted and has two or more adjacent spaces (like in a=b c), the parser would detect an error. This commit fixes that.

Tabs are now also considered like space characters in this case and I added relevant test cases.

Closes #276

[coveralls]

Coverage decreased (-1.07%) to 89.286% when pulling 34d5fe4 on bbc2:fix-unquoted-double-space into 6ca2e2a on theskumar:master.

[x-yuri]

I was reverse-engineering the parsing rules. So I thought, maybe the README can be improved in this PR? Here's what I've managed to find out:

Keys can be unquoted, and single-quoted. Values can be unquoted, single- and double-quoted. Spaces before and after keys, equal signs, and values are ignored. Values can be followed by a comment.

Allowed escape sequences in single-quoted values are: \\, \'.
In double-quoted ones: \\, \', \", \a, \b, \f, \n, \r, \t, \v.

On the other hand, people might start to compose funny env files, which are incompatible with other env file parsers (that mainly concerns ignoring spaces). On yet another hand, env files parsers are incompatible as it is. And well, people might be doing it as we speak :)

[bbc2]

Your reverse-engineering looks correct to me. However, if we're going to describe quoting, there are many other features that we should also describe, like variable interpolation, so I'd rather not do it right now.

On yet another hand, env files parsers are incompatible as it is.

Yes. Before I fully refactored the parser months ago, I had a look at popular parsers for other languages and it's a mess. You can see some of that mess at https://bbc2.github.io/dotenv-parser-comparisons/.

And well, people might be doing it as we speak :)

Indeed they are. It's common to see new GitHub issues created right after a subtle parsing change was made. I've tried to be careful about breaking changes but sometimes you really need them (e.g. when you're the only parser in the world with some weird behavior).