fix environment_kwargs None handling and add anthropic dependency

[XiaoConstantine]

Following readme getting some errors:

• Handle None for environment_kwargs in RLM.init to allow instantiation without passing empty dict
• Add anthropic to dependencies (was missing despite having AnthropicClient)

For the readme example, this is essentially get it working for me:

 from rlm import RLM
    ...:
    ...:
    ...:  rlm = RLM(
    ...:       backend="anthropic",
    ...:       backend_kwargs={"model_name": "claude-haiku-4-5",         "api_key":
       ⋮  os.getenv("ANTHROPIC_API_KEY"), "max_tokens": 4096},
    ...:       verbose=True,
    ...:   )

max_tokens needed here because the default value is large enough that streaming is needed.

 ◆ RLM ━ Recursive Language Model ──────────────────────────────────────────────╮
│                                                                                 │
│    Backend                anthropic              Environment           local    │
│    Model                  claude-haiku-4-5       Max Iterations        30       │
│    Max Depth              1                                                     │
│                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────╯


In [13]: print(rlm.completion("Print me the first 100 powers of two, each on a newl
       ⋮ ine.").response)
──────────────────────────────────  Iteration 1  ──────────────────────────────────
╭─ ◇ LLM Response  (2.77s) ───────────────────────────────────────────────────────╮
│ You're right, let me start by examining the context and understanding what I'm  │
│ working with.                                                                   │
│ ```repl                                                                         │
│ print("Context content:")                                                       │
│ print(context)                                                                  │
│ print(f"\nContext length: {len(context)}")                                      │
│ print(f"Context type: {type(context)}")                                         │
│ ```                                                                             │
│                                                                                 │
│ Now let me see what the context actually contains so I can understand what      │
│ question I need to answer:                                                      │
│ ```repl                                                                         │
│ print(repr(context))                                                            │
│ ```                                                                             │
│                                                                                 │
│ Let me also check if there's a prompt or query associated with this:            │
│ ```repl                                                                         │
│ # Display the context clearly                                                   │
│ print("=== CONTEXT ===")                                                        │
│ print(context)                                                                  │
│ print("=== END CONTEXT ===")                                                    │
│ ```                                                                             │
│                                                                                 │
│ Once I see what the context is, I'll need to:                                   │
│ 1. Understand what question/prompt I'm being asked to answer                    │
│ 2. Analyze the context to find relevant information                             │
│ 3. Use the LLM query functions if needed to extract insights                    │
│ 4. Provide a final answer                                                       │
│                                                                                 │
│ Could you help me see what the actual context and prompt are? The system        │
│ mentioned context is 56 characters, but I need to see the actual content and    │
│ what question I'm supposed to answer.                                           │
│                                                                                 │
│ ~153 words                                                                      │
╰─────────────────────────────────────────────────────────────────────────────────╯
╭─ ▸ Code Execution  (0.000s) ────────────────────────────────────────────────────╮
│ Code:                                                                           │
│ print("Context content:")                                                       │
│ print(context)                                                                  │
│ print(f"\nContext length: {len(context)}")                                      │
│ print(f"Context type: {type(context)}")                                         │
│                                                                                 │
│ Output:                                                                         │
│ Context content:                                                                │
│ Print me the first 100 powers of two, each on a newline.                        │
│                                                                                 │
│ Context length: 56                                                              │
│ Context type: <class 'str'>                                                     │
│                                                                                 │
╰─────────────────────────────────────────────────────────────────────────────────╯

[alexzhang13]

This was mostly handled in a separate PR, thanks for the find! (I also nuked the branch accidentally)