Vellum-ai

- wSDK:
- Fixes node display import on load from module.

- Removes the unused `state` field from `FunctionCall` and `FunctionCallRequest`

- wSDK:
- Fixes a bug when `Json` output type is casted on Templating Nodes.

- wSDK Codegen:
- Fixes a bug with `ChatMessage` values where not all references were being imported as expected.
- Prompts:
- Introduces a new top-level `prompts/` directory where more code-first prompt functionality will be housed.
- Within `prompts/`, introduces a new `compile_prompt_block` helper function that performs all variable
substitutions and templating to generated a compiled prompt block.
- wSDK:
- Introduces a new `nodes/experimental/` directory where experimental node types will be housed.
- Introduces a new experimental `OpenAIChatCompletionNode` that can be used to invoke OpenAI's chat
completion API without needing to route the API call through Vellum.

wSDK Codegen:
- Omits keyword arguments for InlinePromptNode if the value is the same as the default.
- Adds trailing commas to most iterables such that their items are on separate lines.
CLI:
- Introduces a new --include-sandbox flag to the `vellum workflows pull` command that will generate a `sandbox.py` file for easier running of pulled Workflows.

- Includes a number of bug fixes to codegen during the use of `vellum pull`
- Explicitly indicates the default value of the `state` field in Prompt Blocks.