Allow reuse signer across different operators

Applications often create different operators that use identical signers. For instance, databases in the same bucket might share the same OIDC token. Currently, OpenDAL mandates each operator construct its own signer, forcing the application to retrieve a new token every time. This can be inefficient and occasionally cause IAM services to return internal errors.

We should explore methods for reusing signers among various operators instead.

Apart from reusing signer, we should also allow users to configure the cache logic of underlying credentials.

No files, tests, or entry points are identified. Start by locating signer construction and credential-cache handling in the Rust code, then define the reuse and cache-configuration requirements across multiple operators. Done should include coverage for shared signers and configurable underlying credential caching.

Written by the indexing model from the issue text.