fixed load nested config from env

[Sube-py]

When using environment variables to override multi-level configurations, an error will occur if the configuration is of a complex type.

pydantic version: 2.11.9
pydantic-core version: 2.33.2
pydantic-core build: profile=release pgo=false
python version: 3.12.7 (main, Oct 16 2024, 07:12:08) [Clang 18.1.8 ]
platform: macOS-26.1-arm64-arm-64bit
related packages: pydantic-settings-2.8.1 typing_extensions-4.15.0
commit: unknown

Reproduce code:

from pydantic import BaseModel, Field
from pydantic_settings import BaseSettings, SettingsConfigDict


class EmbeddingModel(BaseModel):
    model: str = 'text-embedding-3-small'
    keys: list[str] = Field(default_factory=list)


class LLM(BaseModel):
    embeddings: dict[str, EmbeddingModel] = Field(default_factory=dict)


class LLMSettings(BaseSettings):
    llm: LLM = Field(default_factory=lambda: LLM())

    model_config = SettingsConfigDict(env_prefix='my_prefix_', env_nested_delimiter='__')


# export my_prefix_llm__embeddings__openai__keys='["sk-..."]'
# export my_prefix_llm__embeddings__qwen__keys='["sk-..."]'
print(LLMSettings().model_dump()) # raise error

[hramezani]

Thanks @Sube-py for the PR.
Please add the example in the PR description as a test

[Sube-py]

@hramezani plz check.

[hramezani]

we don't need this comment

Suggested change

	# export my_prefix_llm__embeddings__openai__keys='["sk-..."]'
	# export my_prefix_llm__embeddings__qwen__keys='["sk-..."]'

[hramezani]

please move the test to tests/test_settings.py. you can put it at the end of file. env related tests are there

[hramezani]

@hramezani plz check.

Thanks @Sube-py please check the comments