Thermister -> Thermistor

follow-up the commit b82d91eca6

SERIAL_ECHO is not very good in printing uint8_t - so cast. to int.

Add return to NOT_BUSY for M105
Replace WAIT_FOR_USER with PAUSED_FOR_USER

Add Travis test for FILAMENTCHANGEENABLE

Fix a SNAFU that should have been caught in #3012

Addressing #2773

Partial implementation with reference to MarlinFirmware/Marlin#3011

Alternative fix to #3149

Addressing issue #2767

Aim: Test probes in update_endstops only when activated

Changes:
Configurations
 Add define for FIX_MOUNTED_PROBE to handle the situation where formerly ENDSTOPS_ONLY_FOR_HOMING had to be set, or lowering the nozzle below Z_PROBE_OFFSET_FROM_EXTRUDER could give an "endstop hit" message.
 Add define for Z_MIN_PROBE_USES_Z_MIN_ENDSTOP_PIN to indicate a common situation, that we have a probe but it is connected to an endstop pin
 Add some comments
 Shift some entries to have related things together.

Conditionals.h
 We have a probe (HAS_Z_MIN_PROBE) if one of the pins is defined AND one of the probes is defined.

SanityCheck.h
 Add some tests if the probe is connected and if we have defined a probe.

stepper.cpp
 Changes to test the probe only when it is deployed (z_probe_is_active).
 Test update_endstops() when the probe is deployed.

MarlinMain.cpp
 a. set and reset z_probe_is_active in deploy_z_probe(), stow_z_probe() and dock_sled()
 b. set and reset z_probe_is_active in the case a z-servo is moved to a defined position. The only remaining unhandled servo move is in M280 where we do not end in a defined position. If you want to handle a probe use M401/402
 c. skip deploying/stowing when already deployed/stowed in the dedicated deploy/stow functions.
 d. Handle the new FIX_MOUNTED_PROBE in parallel to a servo driven probe/endstop.

 To do: In another PR. handle all probes in deploy/stow_z_probe.
 Sort out SERVO_LEVELING vs. HAS_SERVO_ENDSTOPS.

If Marlin is blocking the serial input or command queue for any length
of time (for example more than 2 seconds), it needs to send a message
to serial out to inform the host that it is busy. Marlin should only
send these messages out when busy, and preferably not when trying to
print formatted output.

follow-up the commit d56c6af24e

fixes MarlinFirmware/Marlin#3144 and MarlinFirmware/Marlin#3083

lcd_erase_line() is called with argument, but the argument is ignored by lcd_erase_line().

If comparison between signed char(-128 ~ +127) and DIR_NAME_DELETED(uint8_t const 0XE5, it's +229),
it makes incorrect result.

Addition to #3140

diagonal_rod hadn't been used anywhere.