Ignore private, leading _ variables when validate_assignment=True

[maxnoe]

Feature Request

We'd like to configure public (no leading _) variables using pydantic with validate_assignment but still would like private variables _ to be possible, e.g. for logging:

import pydantic
import logging


class ModelWithLog(pydantic.BaseModel):
    option: int = 10

    class Config:
        validate_assignment = True

    def __init__(self, **kwargs):
        super().__init__(**kwargs)
        self._log = logging.getLogger(self.__class__.__name__)  # This fails


m = ModelWithLog()
m.option = 'foo'  # this should fail
m.optoin = 10     # typo, this should fail as well

Result:

Traceback (most recent call last):
  File "model_log.py", line 15, in <module>
    m = ModelWithLog()
  File "model_log.py", line 12, in __init__
    self._log = logging.getLogger(self.__class__.__name__)  # This fails
  File "pydantic/main.py", line 281, in pydantic.main.BaseModel.__setattr__
ValueError: "ModelWithLog" object has no field "_log"```

[samuelcolvin]

duplicate of #655 I'm afraid, please comment there.

[maxnoe]

Ah, sorry did not find this issue somehow.