Set effective_srate to 0 for empty streams when JitterRemoval is off

[cll008]

Currently empty streams crash load_xdf when opts.HandleJitterRemoval = 0, since temp(k).time_stamps is empty and the index is out of range.

[cboulay]

Please also see sccn/xdf#14

What does this change do when streams have no nominal_rate?
(Sorry I don't have Matlab).

[cll008]

Not sure how it affects other downstream functions, but this change does not affect eeg_load_xdf.m since load_xdf.m is called with default options, where opts.HandleJitterRemoval = 1. This change is only relevant when opts.HandleJitterRemoval = 0.

Regardless of the change, a stream without a nominal_srate should crash eeg_load_xdf.m somewhere here since nominal_srate is empty (and effective_srate == 0).

sccn/xdf#14 is actually better though since Jérémy added some warning messages.

[cboulay]

Do you mind testing those changes for me and seeing if it (a) works and (b) fixes your problem? I'd like to merge it but unfortunately I can't test. Actually merging in that repo won't work since we split up the repos, so a new pull request is needed here.

[cll008]

I used xdfBrowser to change a 'healthy' stream's nominal_srate to 0. This file actually did not crash eeg_load_xdf.m since division by 0 in MATLAB results in Inf. eeg_checkset.m addresses nominal_srate ==0 here by changing the sampling rate to 1 and recomputing the timestamps.

I'm not sure the proper git workflow for copying someone's commit when a repo has been split...