MIDI channel view keeps changing when playing MIDI file

[riban-bw]

Describe the bug Playing a MIDI file will cause the MIDI channel selection view to change selection.

To Reproduce Steps to reproduce the behaviour:

1. Start playback of a MIDI file using MIDI Recorder view (enable loop to keep it going throughout test)
2. From main menu, New Synth Chain
3. Select any engine, e.g. FluidSynth
4. Move highlight in MIDI Channel view

Expected behaviour MIDI channel responds to user navigation.

Actual behaviour MIDI channel changes selection based on MIDI messages received from player, making it pretty much impossible to use.

Setup:

• HARDWARE Raspberry Pi 4 Model B Rev 1.4 Audio: ZynADAC Display: ZynScreen 3.5 (v1) Wiring: MCP23017_ZynScreen I2C: MCP23017@0x20, MCP23008@0x21, ADS1115@0x48, MCP4728@0x60
• SYSTEM Raspbian GNU/Linux 10 (buster) Build Date: 2020-09-05 Memory: 37% (703M/1872M) SD Card: 75% (21G/29G) Temperature: 54.5ºC Overclock: None USB Storage: 5% (9.7G/235G)
• MIDI Profile: default.sh Tuning: 440 Hz Receive Mode: Stage (Omni On) ZS3 Sub-SnapShots: On MIDI to Output: Off Master Channel: off
• SOFTWARE zyncoder: quadrant_algorithm (7fd6d35) zynthian-ui: testing (d98a39e) zynthian-sys: testing (af767d2) zynthian-data: testing (cc6a503) zynthian-webconf: testing (90d04eb)

Additional context It looks like there is some kind of helper function zynthian_gui_midi_chan::midi_chan_activity to select MIDI channel from incoming message but this gives poor UX when MIDI channel is changing rapidly, e.g. SMF playing.

[riban-bw]

I believe @wyleu requested this feature. When in MIDI channel list, playing a note on an external controller will jump to that channel in the list. It is a helper to assist with assigning controlelrs to engines / chains but it has a nasty side effect that if other MIDI notes are playing, e.g. from a sequencer, SMF, etc. then the list will jump about like a cat on a hot tin roof. The side effect is significant, sometimes disabling use of the device because navigation is completely broken. I suggest we disable this helper feature (which is undocumented and probably little used) until a better implementation can be applied (triggered by a well though out and structured feature request).