add opus 4.7, adjust thinking

[P3il4]

added opus 4.7 (see its breaking changes)
adjusted opus and sonnet 4.6 to use adaptive thinking instead of the old deprecated budget
set opus 4.7, 4.6 and sonnet 4.6 to 1M context as long context is now priced at standard rates
test:

await puter.ai.chat('What is 127 * 43?', {model: 'claude-opus-4-7', reasoning: {effort: 'medium'}})

sends

...
"thinking": {
    "type": "adaptive",
    "display": "summarized"
},
"output_config": {
    "effort": "medium"
}
...

closes #2812