resolve alt config resolution for jumpstart models

[varunmoris]

Summary

• Fix _build_for_jumpstart() not forwarding config_name to get_init_kwargs(), causing JumpStart models to always use the top-ranked default config
  (e.g. lmi) instead of the user-specified config (e.g. lmi-optimized)
• Add unit tests verifying config_name is correctly passed through the build pipeline

Problem

When deploying a JumpStart model with inference_config_name="lmi-optimized" via ModelBuilder.from_jumpstart_config(), the endpoint used environment
variables from the default lmi config instead of lmi-optimized.

from_jumpstart_config() correctly stores the config name on self.config_name, but _build_for_jumpstart() in model_builder_servers.py never passed it
to get_init_kwargs(). With config_name=None, _add_config_name_to_init_kwargs() falls back to get_top_ranked_config_name(), which returns the default
config.

Fix

Pass config_name=getattr(self, 'config_name', None) to the get_init_kwargs() call in _build_for_jumpstart().