fix: hook onto add_command to propagate errors correctly

[AstreaTSS]

Pull Request Type

• Feature addition
• Bugfix
• Documentation update
• Code refactor
• Tests improvement
• CI/CD pipeline enhancement
• Other: [Replace with a description]

Description

Errors generated when adding prefixed and hybrid commands were not being propagated as they should, as due to being a separated extension, they would error out like a listener instead of like add_command. This caused some fun situations, like the one noted in #1668.

This PR fixes that by adding a way of adding "hooks" onto add_command directly, and then making prefixed and hybrid commands use that hook for its register command logic.

Changes

• Add Client._add_command_hook, which is a list of functions to run during add_command.
• Make prefixed commands and hybrid commands use the hook.
  • This required modifying the register command logic a bit to make the function not async. This would be technically breaking for hybrid commands due to me not using an underscore, so the old function was kept as a wrapper for the new one.

Related Issues

Test Scenarios

@hybrid_slash_command()
@slash_option(
    "test", "The test option", OptionType.STRING, required=True, autocomplete=True
)
async def test(ctx: HybridContext, test: str):
    await ctx.send(f"Test: {test}")

@prefixed_command()
async def test(ctx: PrefixedContext):
    await ctx.send("Hello!")

@prefixed_command(name="test")
async def test2(ctx: PrefixedContext):
    await ctx.send("Hello!")

Python Compatibility

• I've ensured my code works on Python 3.10.x
• I've ensured my code works on Python 3.11.x

Checklist

• I've run the pre-commit code linter over all edited files
• I've tested my changes on supported Python versions
• I've added tests for my code, if applicable
• I've updated / added documentation, where applicable