in case wire reset() somehow allows reading to pass

update dev html-validate@8 and eslint
workaround for form-dup-name (false-positive?)

global espurna::sensor::(anonymous) instead of wrapping every internal one
(pretty sure this is a left-over from incorrectly applying them originally)

ref. #2543
search the bus preemptively and create sensor instances based on that
more clean-up for error handling, allow specific device to fail independently

most of the sensor was re-written
- general one-wire operations are part of the system driver now
- split class definitions for digital and temperature sensors
- system timer instead of polling, enforce ordering by forcing specific
  instance to handle conversion request, while notifying every other ones
- synchronize readings and conversion requests with the sensor reading
  interval, instead of relying on internal polling. now, it happens
  right after begin() and pre() instead of previosly (almost) random tick() calls
- auto-detect conversion time based on device resolution / resolution setting

initialization also adds runtime settings (read on boot)
- `dallasPin` to set current one-wire pin
- `dallasParasite` to change from non-powered mode (default is still on)
- `dallasResolution` to force sensor to be operated using 9...12bit res
  (0 to keep default one)

either timed or instant callbacks, intended to be called from within the
sensor itself or by any other module needing to 'ping' sensor instance

birth message should trigger discovery, enabled by default
ref. #1637

clean-up discovery code to properly re-init when something goes wrong
also restore manual discovery trigger from WebUI

instead of one giant read func, use magnitude-specific ones
ref. datasheet, adc read failure should have a different constant,
and does not warrant re-init

sensor loop assumes value(index) value is ok
ref. #2543

move all checks to pre(), value() returns pre-read value
- tick() and pre() are allowed to fail specific device
- pre() would fail the sensor if any of devices fail
- value(index) is never reached unless tick() and pre() succeed

no real reason to expose internal state, just report filter readiness

ref. #2543

* garland: fix issue with division by zero in first scene setup

* garland: update setters code

* garland: fix web UI

* garland: update scene default settings

* garland: update palettes, add random run animation

* garland: enable and update anim_glow

* garland: update anim_spread

* garland: add anim_crossing

* garland: get rid of redundant variables in main class

* garland: provide anims with all palettes

* garland: unify anim_waves and anim_run to use ColorWave class

* garland: update format for color_wave.h and garland.cpp

* graland: update anim code with clear rands and auto counters

* garland: implement pixel caching for color waves to reduce redundant calculations

* garland: fix anim dolphins out of range issue

* garland: tune anim comets

* garland: combine all wave-based anims; add wave comet anim

* garland: fixes for PR notes

- update library esp8266-oled-ssd1306 to the latest version
- fix thermostat build issues
- simplify and clarify thermostat control logic
- add logs to thermostat control logic

resolve #2596

allow to use api helpers to execute something when button event happens

fix #2592 (for real this time)

cannot declare constexpr objects like that with old compiler
fix #2594

ref. #2585 and 3be53055

fix #2592

re-apply fix from #2181
resolve #2586

minimal & single-source builds should coexist yet again