move setupmodelparameters from each model to abstractNMM

[JohnWangDataAnalyst]

1. move setupmodelparameters from each model to abstractNMM
2. broke example 2 and 5
3. modified the code to make example 2 working (using evaluate method)
4. example 5 (just gpu the new version code incudes the case with multiple blocks, we better use the new version to show how gpu works ) we could just set each par with device as gpu to test it?
5. another reason we should use the new example in zheng_dev for the multimodal is that the examples shown in this repo is fitted to the target value of state E from RWW and did not use the empirical eeg and bold in the cost function, which is actually not a multimodal fitting
6. separating the eeg and bold modules from nmm model is a good idea. In the new version we have those options. modal fitting uses the same script for all the models

[Andrew-Clappison]

This is part of splitting #143 and #144 into smaller PR chunks. Comments also in those closed commits.

Commit 1008fa has a modified ipyb which will add to package size even though it is reverted in a later commit. Modified/reverted in later commits.
It appears the bifurcation objective function component feature which is being used in a paper is removed. Some other objective function simplifications also done.
empRec data object removed, and eg002 content and figures at end commented out/deleted. Content of eg005 commented out or deleted.
Priors added directly in RWWEI2 model class, but in abstract class for other models. RWWEI2 model no longer inheriting from abstracts.
Some transparency from method arguments is removed and made non-optional.

Above is regards to larger changes, changes to equations and code require more detailed review.

[JohnWangDataAnalyst]

will resubmit it