Fix loading encoder weights trained with BYOL

[gaetanbahl]

When loading a model trained with BYOLTask using the "weights" parameter in ClassificationTask, the following error occurs:

Traceback (most recent call last):                                  
File "finetune_ssl_on_eurosat.py", line 14, in <module>                     
    task = ClassificationTask(classification_model="resnet50", loss="ce",
File "/torchgeo/torchgeo/trainers/classification.py", line 103, in __init__
    self.config_task()
File "/torchgeo/torchgeo/trainers/classification.py", line 77, in config_task
    self.config_model()
File "/torchgeo/torchgeo/trainers/classification.py", line 66, in config_model
    name, state_dict = utils.extract_encoder(self.hyperparams["weights"])
File "/torchgeo/torchgeo/trainers/utils.py", line 55, in extract_encoder                       
raise ValueError(
ValueError: Unknown checkpoint task. Only encoder or classification_model extraction is supported 

The extract_encoder function is looking for the encoder key in the loaded dict, when it seems like it should be looking for encoder_name. This PR fixes this and allows to load a model trained with BYOLTask in ClassificationTask for fine-tuning.

[adamjstewart]

It looks like our tests also use "encoder" and need to be updated.

[ghost]

All CLA requirements met.