Fix duplicate locale prefixes in redirect_from paths

[Guflly]

🔤 Polyglot PR

Fixes secondary-language redirect_from paths being prefixed twice during site generation. User redirects remain relative to the language destination, and an explicit matching locale prefix is stripped once.

Closes #313.

Tested with the focused specs, the suite excluding the Unix-only process-count spec, and RuboCop.

Type of change

• Docs update (changes to the readme or a site page, no code changes)
• Ops wrangling (automation or test improvements)
• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Sweet release (needs a lot of work and effort)
• Something else (explain please)

Checklists

• If modifying code, at least one test has been added to the suite

[untra]

Thank you for the contribution @Guflly ! This seems to be reviewing nicely. I'm inclined to include it, however @Guflly it's not clear if you are a person or an automated bot, as you seem to be opening and addressing open issues in many github repos. which is cool; but I need to know there's someone to credit for this work; we need to name a human accountable for work a robot contributes.

Do you have a name or organization you can vouch for, or a multilingual webpage that uses jekyll-polyglot ? Some way to make it clear there is someone real who made this work happen? Something to point to besides your empty github profile that largely suggests you are a sole process that is rapidly contributing to many projects for clout, or otherwise unclear motives.

an AI_POLICY.md document will be added to this project to further clarify this stance. The future is now; and because this project is used by so many I need to be sure all contributions are made by accountable humans or organizations that will take credit and attribution for their work.

[Guflly]

I’m a real person. This PR was AI-assisted: I made and reviewed the code with help from an LLM, as with my recent PRs. I’m accountable for the work.

[untra]

@Guflly I'm sure you are a real person, and you used AI to assist with this work. I'm sure that's really the case.

However you have gone out of the way to make yourself anonymous, and it's unclear who you are, who you work for, and what your aims are. when I look at the commit patch I see you are using + aliasing to make multiple trackable email addresses, and this work comes from 145608489+Guflly@users.noreply.github.com , which cannot be attributed to a named human. or company. or even an email provider. github will likely have an ip address of this committed work, but I bet you went our of your way to commit work from a vpn to hide this too.

For context, this is the worst case scenario I am thinking of as an OSS maintainer of what to prevent:
https://en.wikipedia.org/wiki/XZ_Utils_backdoor

I'm not suggesting you're a nation state actor trying to build arbitrary goodwill across many projects, or that this work is no good. But I am saying that I have seen this go down before, and you are following the threat actor's similar playbook. Hopefully it's to do good by mass contributing help to OSS ; but without another way to identify who you are, I'm hesitant to accept such a contribution.