Add distribution strategy and the ByRemainingRequests implementation

[thekid]

Example

use com\openai\rest\{AzureAIEndpoint, Distributed, ByRemainingRequests};
use util\cmd\Console;

$endpoints= [
  new AzureAIEndpoint('https://...@r1.openai.azure.com/openai/deployments/mini', '2024-02-01'),
  new AzureAIEndpoint('https://...@r2.openai.azure.com/openai/deployments/mini', '2024-02-01'),
];

$ai= new Distributed($endpoints, new ByRemainingRequests());
$payload= [
  'model'    => 'gpt-4o-mini',
  'messages' => [['role' => 'user', 'content' => $prompt]],
];

Console::writeLine($ai->api('/chat/completions')->invoke($payload));
foreach ($endpoints as $i => $endpoint) {
  Console::writeLine('Endpoint #', $i, ': ', $endpoint->rateLimit());
}

[thekid]

Released in https://github.com/xp-forge/openai/releases/tag/v0.4.0