bot: Support multiple handlers for a URL pattern

[half-duplex]

Description

Implements #1804

This introduces Sopel._url_callbacks, changes the *_url_callback functions to use it, and returns both Sopel._url_callbacks and bot.memory["url_callbacks"] results in a search, maintaining the legacy functionality.

Because Sopel doesn't have a SopelMemory-type object for lists, this may have minor concurrency issues, but url_callbacks updates are infrequent enough I don't think it's deadly.

Checklist

• I have read CONTRIBUTING.md
• I can and do license this contribution under the EFLv2
• No issues are reported by make qa
• I have tested the functionality of the things this change touches

[Exirel]

I like the idea, and the code is quite good. Yet, I'd like to point out that we planned on removing memory['url_callbacks'], to prevent any manual access to this feature - as mentioned in the issue:

At the same time, we can officially remove the memory dict.

For that, I think you can add a new attribute to Sopel, something like Sopel.url_callbacks maybe? It could be a SopelMemory object, where keys are compiled regex. Alternatively, it could be a protected attribute, like _url_callbacks.

[half-duplex]

Thanks. Yeah, this was originally aimed for 7.0 which needs to keep manual management. Mostly redone now to be fully compatible with no API changes at all, so this maintains exact compatibility with the manual memory["url_callbacks"] way.
I like it more as a private attribute, since I can't think of any valid reason to directly work with it from a plugin, and when private we don't need to worry about compatibility if it requires any further changes.

[Exirel]

I like it. Few nitpicks to consider.

[dgw]

I'm in favor of doing this for 7.1. Where are we on addressing Exi's feedback?

[Exirel]

Can we keep the line here? I like the code with a bit of space, for readability.

[dgw]

I think it was to make clear that the whole block should be removed in 8.0 per the above comment. I'm fine with or without the blank line here, but I understand the (suspected) reasoning behind deleting it.

[Exirel]

You should check if callback is not yet registered for that pattern.

[Exirel]

You might want to catch a KeyError here too.

[dgw]

I suspect #1904 will replace this. Even if not, this will need some refactoring after the recent massive changes to how plugin rules work. @Exirel confirmed so on IRC.