Document design of dial assistant context management

[avavilau]

• Target state (design)
• API Spec for Assistant limits
• Smaller task list with ETAs

[Oleksii-Klimov]

Target state:

• Assistant provides prompt truncation/tokenization endpoints to ensure that completion requests can fit: Implement tokenization/truncation endpoint #38
• Completion request provides an extra parameter to control how many tokens can be dedicated for model dialogue with addons:
  Add max_addons_dialogue_tokens parameter to the completion request ai-dial-sdk#51
  Support max_addons_dialogue_tokens parameter from completion request #40

Usage scenario:

1. Tokenize the completion request using the tokenization endpoint.
2. If the request fits then go to step #⁠4.
3. If the request does not fit, reduce the number of messages or message sizes then repeat step #⁠1.
4. Send the request to the chat completion endpoint.
   The truncation endpoint aids in managing history size by allowing for the discarding of old messages.