implement clientside signal processing

[bcliang]

Ideas

• use session store for input buffer (vs. recalculating trace each time)
• simple moving average (Math.mean)
• kalman filtering (kalmanjs)
• generate new trace (add/extend) for filtered data on plot

[bcliang]

some working code exists here https://github.com/bcliang/dash-clientside-demo/tree/add-clientside-filter-example

TODO: because traces can have non-matching length (filter enabled/disabled) need a way to output CSV with aligned time. ALTERNATIVELY, just draw each trace as a separate pair of (X,Y) columns

[bcliang]

Made changes -- each trace has separate (x,y) paired column in CSV.

TODO: Refactor to eliminate confusing input validation performed in ui.relayout() and signal.filterSignal