Feature/pcf8591

[gen2thomas]

this PR will cover the issue #813

Adds in detail

• possibility to use AnalogSensorDriver for scaling values to float64
• add possibility to switch off polling with given 0s as polling intervall, although Start() is called
• add generic TemperatureSensorDriver which supports nonlinear NTC, PT100, PTC and other linear sensors
• adjust grove temperatur sensor to the generic one, add tests for correct scaling the NTC with B=3975, R25=10k, Rs=10k
• introduce the AnalogWriter interface
• introduce AnalogActuatorDriver which uses the new interface
• add driver for PCF8591 to cover support for PCF8591 8-bit 4xADC 1xDAC i2c #813
• add driver for YL-40 module which contains the PCF8591, temperature sensor (B=3950, R25=10k, Rs=1k, invers), brightness sensor, potentiometer, free analog input, one analog output, use all previously added features
• examples for PCF8591 and YL-40 module provided
• most features practically tested with tinkerboard

Notes for NTC scaling

• the scaling can be done by T0, R0, B or also by T0, R0, T1, R1 --> B will be calculated at initialization
• thermistor is typically connected with an series resistor between reference voltage and ground, both directions are supported (normal, inverse)
• all NTC test expectations were calculated by an LibreOffice table

Notes for the PCF8591 chip and the driver

The IC PCF8591 only supports 100 kHz (basic 100 kbit/s mode), see section 14.4 Dynamic characteristics at page 20 of NXP specification. Running the PCF8591 in 400 kbit fast mode, which seems to be the default of most platforms now, leads to some instability (tested with 5 devices at ~370 kbit/s on tinkerboard):

1. the control byte is not recognized on first write (on 4 devices, ~30% of calls)
2. the read value is to small when immediately read after write (all devices, 100% of calls)
3. the read value is still the old one after more than one reads (all devices, ~99% of calls)
4. the read value is still the old one after more than two reads (4 devices, ~50% of calls)

strategies to overcome problems with 400 kbit:
for 1.: do "full double reading" (write control register and read, done in most other drivers) fix the problem for 3 tested devices, ~95% good calls
for 3,4.: multiple write don't stabilize this enough (only reduce to ~75%), multiple reading to an buffer and after some relaxation a single read is quite sufficient to reduce it to 5% wrong reads (for one device the problem is gone completely)
for 2.: for long conversion time, write also don't do anything for stabilization, it will be covered by the strategy above

driver:

• all modes are supported (single, 2-/3- differential, mixed) - most other drivers (python, cpp) don't do that
• auto increment is not supported yet
• 400 kbit stabilizing strategies are part of the driver (default is 2x write, 3x buffered read, wait, 1x read)

Notes for YL-40 module driver

Because the supply voltage of i2c bus on tinkerboard seems to be 3.3V and both LED's are shining very bright, I would suggest to not run this board with 5V. Therefor some scaler are prepared for 3.3V as default. This can be changed by a with-function to each scaler function you want.

[deadprogram]

Another massive PR, you have me in awe @gen2thomas

Merging 🎆