log nparray() drops entries with duplicate time points.

[davidt0x]

Calls to nparray() on log don't report entries with duplicate time values. This is inconsistent with the report generated by print_entries().

Reproducible example from @jongkeesbj

import psyneulink as pnl

my_lca = pnl.LCAMechanism(
        size=2,
        function=pnl.Logistic(gain=1.0),
        termination_measure=pnl.TimeScale.TRIAL,
        termination_threshold=10,
    )

my_lca.set_log_conditions([pnl.RESULT])

comp = pnl.Composition(my_lca)

comp.run([1, 0])

my_lca.log.print_entries() # This shows results for all 10 time steps
my_lca.log.nparray() # This shows results only for the last time step?

This issue seems to be caused by nparray() using Log._assemble_entry_data which drops duplicate entries with the same time. I think the correct fix would be to add a further level of granularity to time that includes the execution number.

[jdcpni]

Have implemented a record_all_executions option in log.nparray() and log.nparray_dictionary(). Setting this to True causes the values of all executions within a TIME_STEP to be reported as a list for that entry in the log. Note that this adds an extra list dimension to the values as compared to when the option is False.

[kmantel]

Fixed by #2844