google-auth: gracefully handle leading whitespace in identity pool PEM parsing

### Summary of the issue
`_read_trust_chain` currently splits the PEM file on the literal `-----BEGIN CERTIFICATE-----` string.

If a perfectly valid PEM file happens to have some comments or blank lines at the top, that garbage ends up in the first split element. This causes `cryptography` to throw a `ValueError` and crash when it tries to load the chain.

### Proposed fix
Instead of a brittle string split, we should use `cryptography`'s native multi-PEM parser so it handles the noise gracefully.

### Affected Files
* `google/auth/identity_pool.py`

Start in google/auth/identity_pool.py at _read_trust_chain and inspect how the PEM content is split before cryptography loads the certificates. Done means valid PEM files with leading comments or blank lines are parsed without a ValueError or crash, while the existing certificate chain behavior remains intact.

Rédigé par le modèle d'indexation à partir du texte de l'issue.