[BUG] tensorboard does not work with gridsearch when model_name is provided

[gdevos010]

Describe the bug
providing a model name when using gridsearch puts all files under the same folder ./darts/runs/MODEL_NAME
this breaks tensorboard

To Reproduce

NBEATS_params = {
        "log_tensorboard": [True],
        "model_name": ["NBEATS_gs"],
        "layer_widths": [256, 512, 1024],
}

best_model, dict_params = NBEATSModel.gridsearch(NBEATS_params, train_air, val_series=val_air, metric=mape, n_jobs=3, verbose=True)

Expected behavior
each run should have its own dir

System (please complete the following information):

• Python version: 3.7
• darts version 0.16.0

Additional context

[hrzn]

Thanks for flagging. Indeed currently gridsearch was not really meant to work with Tensorboard. Anyway we have a bit of refactoring to do around gridsearch. Leaving this one open for the time being.