Fix missing log steps at 0.1s period

[mha1]

Fixes #1966

Summary of changes

• simulator: allow for one get_tmr10ms() tick (10ms) deviation
• radio: switched to FreeRTOS software timer at user requested period
• code cleanup:
  log.cpp: made variable lastLogTime static
  log.cpp: deleted unused variable g_logError
  functions.cpp, logs.cpp, sdcard.h: renamed variable logDelay to logDelay100ms to show it is scaled to 100ms increments

Thanks to @rotorman!

[mha1]

Here's a little before/after comparison. Setup is a TX16s with a non-trivial model selected and logging enabled at 0.1s for 15 minutes. The radio is not touched, just sitting there logging.

Before: current 2.7.1 implementation showing a log with 8833 log steps (roughly 15 minutes). It can be seen that the expected dt of 100ms is not well maintained with more than 12% off the requested 100ms.

 dt     cnt    freq

1 100 7735 0.876
2 110 545 0.062
3 140 545 0.062
4 150 5 0.001
5 80 1 0
6 120 1 0
7 230 1 0

After: current main with using software timers as proposed in the PR in the same setup as above with 9071 log steps at 0.1s. dt's are locked to 100ms. The single 80ms dt is the first log step entry.

 dt      cnt     freq

1 100 9070 1
2 80 1 0

I tried different scenarios but would appreciate community testing

[mha1]

@raphaelcoeffic - can please review this pr, thx

[raphaelcoeffic]

Looks good to me! Thx!

[mha1]

@raphaelcoeffic Thx and sorry for bugging you

[raphaelcoeffic]

@raphaelcoeffic Thx and sorry for bugging you

No problem! That was the right thing to do! I have quite some review work to catch up. The LVGL stuff has literally sucked up my time…