Fix publish filename resolution

[stevenjoezhang]

What does it do?

This PR fixes draft filename resolution used by hexo publish:

• Accepts exact filenames, including extensions such as .md and .mkd.
• Accepts paths relative to _drafts, including nested directories.
• Allows the extension to be omitted when the filename is unambiguous.
• Prevents a filename such as foo from incorrectly matching foo-bar.md or foobar.md.
• Reports an explicit error with the available candidates when multiple files have the same stem.
• Handles filenames containing spaces.
• Normalizes path separators and respects the filename_case configuration.
• Separates source-file lookup from the slug used to create the destination post.

The destination filename continues to follow the existing new_post_name behavior.

This PR intentionally does not include the unpublish command or implementation. That work remains in #4995 so the original contribution and commit history are preserved.

Fixes #1821.
Supersedes #5476.

Screenshots

N/A — this change only affects CLI behavior.

Pull request tasks

• Add test cases for the changes.
• Passed the CI test.

[github-actions]

How to test

git clone -b fix/publish-unpublish-filename-resolution https://github.com/hexojs/hexo.git
cd hexo
npm install
npm test

[github-actions]

Flamegraph

https://9009d0b7a6b816af69163df663750d59771b43f2-20-hexo.surge.sh/flamegraph.html
https://9009d0b7a6b816af69163df663750d59771b43f2-22-hexo.surge.sh/flamegraph.html
https://9009d0b7a6b816af69163df663750d59771b43f2-24-hexo.surge.sh/flamegraph.html

[yoshinorin]

I think the original contribution should be preserved here. The description references #4995 by @awongCM in prose, but all commits are authored by you, so their contribution doesn't appear in the git history. When a change builds on an existing implementation, that attribution should be carried in the commits themselves, not only referenced in the PR description.

Given that, I think one of the following should happen:

1. Have @awongCM recreate the work.
2. Preserve @awongCM's commits during the rebase (though I suspect this was difficult given how old the original PR is).
3. Get @awongCM's approval to merge this PR as is.

[stevenjoezhang]

Thanks for pointing this out. I agree that the original contribution and its Git history should be preserved.

I have split the changes and force-pushed the PR on top of the latest master. #5805 now contains only the publish filename-resolution fix for #1821/#5476. The unpublish command, implementation, and tests have all been removed and remain in #4995.