Make use of the data access role for the tiler permissions instead of a bucket list

Right now, the titiler pgstac construct has a `buckets` parameter, and this list of buckets is added to the lambda execution role at deployment time.

This is inconsistent with how we handle permissions for the ingestor lambda : we require the user to provide the ARN of a pre-existing 'data access role'.

Because the ingestor and the titiler lambdas will be accessing the same assets, it makes sense to configure these lambdas with the same permissions.

Handling these permissions in a separate, reusable role (that a user potentially reuses in other applications than the tiler/the stac ingestor) is preferable to avoid having duplicate definitions of the same permissions.

We should therefore parameterize the tiler construct with the same `data_access_role_arn` option.