feat: align Moonshot model IDs, reasoning-history guidance, published README, documentation, and examples (#19581)

## Background

Moonshot consumers need editor completions, examples, and published
guidance that match the current Kimi Chat Completions catalog while
retaining support for custom and retired model IDs.

## Summary

Adds first-class moonshot-v1-auto and three Moonshot V1 vision-preview
model IDs while preserving the arbitrary-string escape hatch. It keeps
budgetTokens as a deprecated compatibility field, documents actual
reasoningHistory behavior, and preserves existing runtime request
handling.

## Testing

Compile-time coverage verifies the complete 12-ID first-class union and
arbitrary or retired strings. Runtime coverage constructs each newly
typed model. Moonshot Node and Edge suites pass with 99 tests each, and
package, example, and full-workspace type checks pass.

## End-to-end Validation

- Successfully exercised live Kimi K3 reasoning effort, Kimi K2.6
thinking, and Kimi K2.7 Code preserved-thinking examples.
- Confirmed all typed IDs, multimodal capability flags, and Kimi K3
low/high/max reasoning metadata through /v1/models.

## Documentation

Updates the Moonshot provider reference, provider index, ai-functions
examples, and published package README with current recommended models,
K3 reasoning effort, K2.6 and K2.7 thinking guidance, accurate
reasoningHistory semantics, Moonshot V1 vision and sunset information,
and current Kimi platform links.

## Related Issues

Fixes #19555

---------

Co-authored-by: ai-sdk-factory <308175966+ai-sdk-factory@users.noreply.github.com>
Co-authored-by: gr2m <39992+gr2m@users.noreply.github.com>