Implement ExperimentSpec to reduce redundancy

A lot of the code/experiments needs access to static information such as:

• device name
• experiment name
• subject id
• session id
• data directory
• record duration

It would probably be a good idea to add a dataclass for this, that'll be easier to pass around without having to repeat the same set of arguments every time.

The idea would be that once you've constructed a ExperimentSpec, you can:

• Run the experiment: expspec.run()
• Get an EEG instance: expspec.device (generated from input arguments at initialization)
• Generate a savefile: expspec.generate_output_file() (see #74)
• Write new experiments by simply having their present functions accept the single argument spec which contains all metadata possibly needed by the experiment.

Essentially:

@dataclass
class ExperimentSpec:
    device_name: str
    experiment_name: str
    subject_id: int
    session_id: int
    device_options: dict = field(default_factory=dict, ...)

    def __post_init__(self):
        self.device = EEG(self.device_name, **self.device_options)

    def run() -> None:
        ...

    def get_output_file() -> Path: 
        ...

Start by tracing how existing experiments receive device, metadata, and output-file arguments, then inspect how EEG instances are constructed and read issue #74. Done means a coherent ExperimentSpec API supports experiment execution, device access, and output-file generation while existing experiment entry points can use the single spec argument.