This also replaces instances where we manually write "return ctx.Err()"
with "return context.Cause(ctx)" which is functionally identical, but
will also correctly propagate cause errors if present.
Currently, the Session proto id_token field is populated with Pomerium
session data during initial login, but with IdP ID token data after an
IdP session refresh.
Instead, store only IdP ID token data in this field.
Update the existing SetRawIDToken method to populate the structured data
fields based on the contents of the raw ID token. Remove the other code
that sets these fields (in the authenticateflow package and in
manager.sessionUnmarshaler).
Add a test for the identity manager, exercising the combined effect of
session claims unmarshaling and SetRawIDToken(), to verify that the
combined behavior is preserved unchanged.
