fix: revert max_completion_tokens to max_tokens (#1741)

lua/avante/providers/azure.lua

@@ -2,7 +2,7 @@
 ---@field deployment string
 ---@field api_version string
 ---@field temperature number
----@field max_completion_tokens number
+---@field max_tokens number
 ---@field reasoning_effort? string
 
 local Utils = require("avante.utils")

lua/avante/providers/openai.lua

@@ -67,8 +67,6 @@ function M.set_allowed_params(provider_conf, request_body)
   else
     request_body.reasoning_effort = nil
   end
-  -- If max_tokens is set in config, unset max_completion_tokens
-  if request_body.max_tokens then request_body.max_completion_tokens = nil end
 end
 
 function M:parse_messages(opts)