feat: apply pretty_url to url_for & full_url_for

[SukkaW]

I have noticed url_for & full_url_for will not remove index.html while trailing_index is disabled. This PR is addressed to make the trailing_index behavior more consistent.

I haven't apply pretty_url options to relative_url() yet because relative_url doesn't require bind(hexo) and can not read hexo.config directly. After we finish relative_url() I will make this PR "Ready to Review".

[coveralls]

Coverage increased (+0.02%) to 95.302% when pulling 07705cd on SukkaW:pretty_url into beb8526 on hexojs:master.

[yoshinorin]

url_for & full url_for implementations are LGTM.

[curbengh]

LGTM too.

[curbengh]

how about adding an optional boolean parameter to relative_url()?

relative_url(from, to, true), then remove trailing index.html if false.

I'm assuming ../../ is equivalent to ../../index.html?

[curbengh]

Due to urgency of 1.6, I don't mind holding off relative_url to later.

[SukkaW]

@curbengh pretty_url will have trailing_index, trailing_html, maybe even more options in the future. A simple boolean parameter just won't help.

[curbengh]

pretty_url will have trailing_index, trailing_html, maybe even more options in the future.

what about codeblock() tag plugin approach? e.g. relative_url(from, to, { trailing_index: true, trailing_html: true }).

Another idea I just thought of is that making bind(hexo)/call() optional, where those functions can be used as url_for(input, siteConfig). They can check existence of this.config.url and siteConfig.url and respond appropriately.

This is useful where this object doesn't refer to hexo (like in hexo-renderer-marked) or the hexo object can't be accessed directly.

[SukkaW]

I think pass a config as the last parameter could be a good idea. We first check if that parameter is given, if doesn't, then get config from this.