Context

Vapi's generated SDK was previously having memory issues on import, so we had implemented a temporary workaround to custom write the __init__.py files.

Changes made

• Regenerated SDK on the latest fern generator version which includes memory improvements
• __init__.py and client.py files now implement lazy loading
• Removed __init__.py files from fernignore

Memory Improvements

Initial testing is showing a 50% reduction in initial RSS memory on load. (vapi_server_sdk-1.7.1-py3-none-any.whl is the package for this PR, and vapi_server_sdk-1.7.0-py3-none-any.whl is the existing version)

Package: vapi_server_sdk
Import statement: from vapi import Vapi
Total wheels analyzed: 2

Wheel                                              Size (MB)    RSS Diff (MB)  RSS Before   RSS After   
# original version
vapi_server_sdk-1.7.0-py3-none-any.whl             0.91         49.64          14.45        64.09          

# this PR 
vapi_server_sdk-1.7.1-py3-none-any.whl             1.10         24.33          14.98        39.31