Fix optional embedded model

[gsakkis]

Supersedes #273 in fixing #272.

[SYNchroACK]

@art049

[torrell8]

Hello, I am experiencing the same problem.
Is there an update on this merge request?

[yutongp]

@art049 this is quite essential feature. I ran into this issue the first time I use odmantic. Is there any update for a merge?

[Cajuteq]

@art049 Sorry for pinging you but this PR seems quite compact to review and would resolve quite a lot of (mostly redundant) issues.
So far I have found Issues #272 #350 #370 #477 #484 and PR #273 #344 #485 to be somehow related.
Maybe you could take a look.

[WolfgangWenzel]

It is really sad that his is not fixed but i think there is a workaround which works for some use-cases: make the embedded model not optional and write a model-validator 'before' which adds the field with some default values:

    @model_validator(mode='before')
    @classmethod
    def add_slurm_parameters(cls, data):
        if not 'slurm_parameters' in data or data['slurm_parameters'] is None:
            data['slurm_parameters'] = SlurmParameters()
        return data

here slurm_parameters (which is the embedded model) is created with some defaults when not present.