QAT range learning tracker

[andrewor14]

This issue trackers follow-up feature requests regarding QAT range learning after the initial PR: #2033

• Do not learn zero points for symmetric quantization
• Add option to learn actual ranges instead of scales and zero points directly
• For weight-only quantization, allow users to initialize the new parameters in __init__ instead
• Allow users to save learned scales and zero points and reload them in a new training job