refactor batch item file management

[kushalkolar]

Refactor all algos and extensions so that output data that corresponds to a batch item are stored within their respective subdirs within the batch dir (which is the same dir as a batch file). Raw input movies remain where they are.

Created PathsDataFrameExtension and PathsSeriesExtension in batch_utils module to simplify resolving or splitting paths of input and output files for batch items.

The following extensions will automatically resolve the full path, first searching within the batch dir and then within the parent dir of the current parent_raw_data_path.
df.paths.resolve(<path>), also works for Series for example: df.iloc[0].resolve(<path>

This functionality require the current main branch of pandas as of June 19, 2022. Specifically it requires that the dict pd.DataFrame.attrs gets propagated to child pd.Series.attrs. This functionality will apparently be released in pandas v1.5.0 on June 30, 2022. For more info see: pandas-dev/pandas#46101

Example:

load_batch(<batch_path.pickle>)
batch_item = df.iloc[-1]
input_movie_path = batch_item.params["input_movie_path"]
full_input_movie_path = batch_item.paths.resolve(input_movie_path)

also df.caiman.run() now uses subprocess backend by default and doesn't required the batch_path as a kwarg, it gets it from df.paths.get_batch_path()

[kushalkolar]

WIP, tests pass on my end for mcorr, working on CNMF and CNMFE