Allow non-kebab-case names in component ai_models

[seun-ja]

Aims to resolve #3256

Relaxes rules to accommodate LLM models. Also includes tests and updates current ones

[vdice]

I'm not sure if this is the ideal way to go about the fix. Here's my initial thought:

With the current version of this PR, all keys/values in the manifest are now exempt from the exceptions added here. This is problematic specifically for Spin application names (and maybe for other keys; I haven't yet looked in depth). We have logic elsewhere (eg spin new and spin deploy) with more restrictive naming rules, which I think we want to continue to uphold. For instance, admitting the : character in an app name would cause its OCI artifact reference to no longer be valid, and spin deploy would fail.

I'm not quite sure about the genesis of forbidding a segment from starting with a number, though @lann may recall. It's possible that is a restriction that could be lifted...

Hopefully @lann (or @itowlson?) can advise here as I'm still familiarizing myself with the code. My naive thought was we'd update logic (either in the serde crate here or maybe the toml parser?) to specifically ignore ai_models string values. Or, if ignoring is too lax, basically just apply the exceptions you've added here to these string values, but continuing to preserve the default restrictions for everything else in the manifest.

[fibonacci1729]

At the moment, I don't see much value in restricting the IDs of ai_models to be kebab-cased (iirc this was put into place in anticipation of value imports in the component model which won't land for a longish time). I think we can just use String in lieu of KebabId here and remove the compat enforcement here.

[itowlson]

@seun-ja Unless there's value in having the history (or it's hard to do) I'd suggest squashing the commits here.

[vdice]

Nice, love how simple this fix turned out to be! If it is a quick add, should we add unit test coverage?

[fibonacci1729]

Looks good to me! Maybe just update the title to reflect the new changes.

[itowlson]

@seun-ja Could you GPG sign please? Thanks!

[seun-ja]

@itowlson GPG signed.

Out of curiosity, could it be because I squashed my commit that could have made it seem like I didn't GPG sign? The commit even passed the DCO check

[itowlson]

@seun-ja Oh, that's possible. Squashing would likely have retained the signed-off-by text which is the DCO, but if you're manually GPG-signing then it's a new commit and so presumably waves hands a new cryptographic thingummy needs to be whatsited.

If you want, you can turn on automatic GPG signing by setting commit.gpgsign=true in your git config - I have this and so I never have to think about whether I need to re-sign after a merge/rebase.

Anyway thanks for doing it, I see the Verified flag and I've queued an auto merge.