Fido: Fix COSE decoding with x-only keys

[p1gp1g]

Some EC keys uses only the x-coordinate (cf. https://datatracker.ietf.org/doc/html/rfc9053#name-elliptic-curve-keys), and used to crash while trying to GetByteString.

This fix the Cbor decoding for these keys

[peterhel]

Nice fix — the runCatching + nullable y lets x-only / OKP keys decode instead of throwing in GetByteString(). Builds clean here on current master.

One follow-up I hit while reading it: asCryptoKey() still does

ECPublicKeySpec(ECPoint(BigInteger(1, x), BigInteger(1, y)), parameterSpec)

so a key that now decodes with y == null will NPE there (BigInteger(1, null)) rather than failing cleanly at decode time. It pairs naturally with #3545, which reworks key construction for Ed25519/OKP — so either landing them together or guarding asCryptoKey() for the null-y case would close that gap. Happy to test against a real x-only assertion if useful.

[p1gp1g]

I should have push both in a single PR indeed