Fix openai#embed method

[andreibondarev]

No description provided.

[andreibondarev]

@prdanelli Would you please give this a proper review and test?

[prdanelli]

Andrei, I can confirm it works as intended - we are able to specify the dimensions as a param to the embed method:

irb(main):003> client.llm
=>
#<Langchain::LLM::OpenAI:0x0000000113ef69b0
 @client=
  #<OpenAI::Client:0x0000000113ef6348
   @access_token="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",
   @api_type=nil,
   @api_version="v1",
   @extra_headers={},
   @faraday_middleware=nil,
   @organization_id=nil,
   @request_timeout=120,
   @uri_base="https://api.openai.com/">,
 @defaults=
  {:n=>1,
   :temperature=>0.0,
   :chat_completion_model_name=>"gpt-3.5-turbo",
   :embeddings_model_name=>"text-embedding-3-small",
   :dimension=>512}>

irb(main):004> vs = client.llm.embed(text: "cat").embedding
irb(main):005> vs.count
=> 1536
irb(main):006> vs = client.llm.embed(text: "cat", dimensions: 512).embedding
irb(main):007> vs.count
=> 512
irb(main):008> vs = client.llm.embed(text: "cat", dimensions: 999).embedding
irb(main):009> vs.count
=> 999

It looks as though the default is no longer respected when the LLM instance is created:

Langchain::LLM::OpenAI.new(
      api_key: ENV["OPEN_AI_SECRET_KEY"],
      default_options: {
        embeddings_model_name: "text-embedding-3-small",
        dimension: 512,
      }
    )

But that might be one for another day. Thank you again for jumping on this so quickly, its very much appreciated - I'm sorry I wasn't able to add the PR myself.