Allow `signer=None` in client initializations

The client initialization code assumes that, if signer appears in the keyword arguments, then it's a valid signer.

This prevents the use of a generic setup that either allows a config or a signer to be passed.

For example, this code does not work with the current version:

    if USE_INSTANCE_PRINCIPAL:
        config = {}
        signer = oci.auth.signers.InstancePrincipalsSecurityTokenSigner()
    else:
        config = oci.config.from_file()
        signer = None

    identity = oci.identity.IdentityClient(config=config, signer=signer)

Instead, you have to duplicate the initialization for signer based on the code that exists in each of the client class initializers.

    else:
        config = oci.config.from_file()
        signer = oci.Signer(
                    tenancy=config["tenancy"],
                    user=config["user"],
                    fingerprint=config["fingerprint"],
                    private_key_file_location=config.get("key_file"),
                    pass_phrase=oci.config.get_config_value_or_default(config, "pass_phrase"),
                    private_key_content=config.get("key_content"),
                )

Which is fine, until Signer changes in a future version of the SDK and this breaks. Plus, code duplication should never be required :(

Start with the client class initializers and the IdentityClient initialization shown in the issue, then trace how config and signer are validated. Confirm the intended behavior for signer=None across client initializations and add coverage showing that both signer-based and config-based setup continue to work.