fix(vchordg): avoid panic when validating an empty alpha option

[ameyypawar]

What

VchordgIndexOptions::validate_alpha indexes alpha[0] after the is_sorted() and range checks — both of which are vacuously true for an empty slice:

fn validate_alpha(alpha: &[f32]) -> Result<(), ValidationError> {
    if !alpha.is_sorted() { ... }                            // [].is_sorted() == true  -> skip
    if !alpha.iter().all(|x| (1.0..2.0).contains(x)) { ... } // [].iter().all(..) == true -> skip
    if alpha[0] != 1.0 { ... }                               // [] -> index out of bounds
    Ok(())
}

So an empty alpha panics (index out of bounds) instead of returning a ValidationError. It's reachable from SQL via WITH (options = '[index] alpha = []'), which surfaces a cryptic panic message rather than a clean validation error.

Fix

Use alpha.first() so the empty case yields the existing `alpha` should contain `1.0` error:

if alpha.first() != Some(&1.0) {
    return Err(ValidationError::new("`alpha` should contain `1.0`"));
}

For any non-empty input alpha.first() == Some(&alpha[0]), so behavior is unchanged; only the empty slice goes from panic to a proper validation error.

Tests

Adds unit tests for validate_alpha covering the empty slice (regression) plus valid/invalid cases. These run under the existing cargo test --workspace --exclude vchord job (vchordg is a pure-Rust crate).

Happy to adjust the wording or drop the tests if you'd prefer a more minimal change.

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}