G29 E was not retracting the probe on the last probe point when used with ABL grid. Corrected.

* Adds config parameter `PID_PARAMS_PER_EXTRUDER` - allows single PID
parameters to be used where this would be preferable (e.g. dual
identical extruders)
* When disabled, will use `float Kp, Ki, Kd, Kc;` as before.
Preprocessor macros used to switch between.
* ultralcd.cpp defines extra menus for extra parameters only where
required
* M301 reports `e:xx` only if independent pid parameters enabled
* EEPROM structure still leaves space for 3 extruders worth, when undef
will save single parameter to all extruder positions, but only read the
first
* Switching off saves approx 330 B with no LCD enabled, 2634B with LCD
(RRD) enabled: this is significant.
* LCD modifications should be tested.

This change introduces an improvement to G29 command on Marlin.

Auto bed leveling operation's reliability is based on the repeatability of the Z-probe switch and the servo. This change introduces an option to G29 command. When the G29 command is sent with an "e" option, during auto bed levelling the servo is not retracted between probes which decreases the bias on auto bed levelling resulting from servo.

G29 command does the auto bed probing as it is.

G29 E command engages the servo on first probing point, probes all points and retracts the servo after probing the last point.

Please comment your opinions, test on your printer and check the code on a programmer's perspective. (I am not a good programmer.)

* Variables Kp, Ki, Kd, Kc now arrays of size EXTRUDERS
* M301 gains (optional, default=0) E parameter to define which
extruder's settings to modify. Tested, works with Repetier Host's EEPROM
config window, albeit only reads/updates settings for E0.
* All Kp, Ki, Kd, Kc parameters saved in EEPROM (version now v14), up to
3 extruders supported (same as Marlin in general)

try, try again.

If 'OK' is garbled on sending PC won't receive it.  Both machines will
wait on each other forever.  This resends OK if nothing is heard from PC
for a while to avoid this bad case.

fixed minor typo

void manage_inactivity(bool ignore_stepper_queue=false)

standard is false so it is not necessary to change additional code. this
change should be better, hopefully nophead like this :)

more speaking name

make a more general solution

Tested for SAV MKI and SAV 3D LCD on lewihe.

- Fixed issue when BAL area probing is shorter than it should be for
grid probing
- Warning when BAL activated with Delta Kinematics
- Fix XY_TRAVEL_SPEED when homing Z axis

Changed level and leveling to compensation (except with "verbose_level" and "lcd_reset_alert_level").

in

G0 to G3 were previously acknowledged in the get_command method, causing
them to be possibly acknowledged before commands coming after them that
were acknowledged in process_commands.

This patch fixes this, moving the acknowledgement of G0 to G3 to the
process_commands method as well. These commands are therefore no
longer acknowledged when the enter the cmd_buffer but instead only
acknowledged when the enter the plan_buffer.

Guaranteeing that commands are acknowledged in the same order in which
they were received by the firmware allows host software to be able to
track the life cycle of commands and such a better management of the
firmware's serial buffers as well as better internal command processing
and response parsing without having to depend on throwing an epic
amount of regular expressions against each line received back from the
firmware.

Fixes ErikZalm/Marlin#1147

The welcome message is more appropriate than “resuming print” if
nothing is printing.

Allow M0 and M1 to include a message string. Ignore clicks on “Wait for
user” so that the Info Screen stays up.

The welcome message is more appropriate than “resuming print” if
nothing is printing.

Allow M0 and M1 to include a message string. Ignore clicks on “Wait for
user” so that the Info Screen stays up.

This function should only be used with sled, not all bed auto
levelling.

Improvement to avoid reinitializing delay buffer with every print. Fixed
issues in buffer indexing and memory out of bounds due to floating point
imprecision.  Simplified the code by avoiding conversion to standard
diameter and 1cu mm extrusion, which caused complications in determining
mm extruded.

chdkActive was set to false regardless of (millis() - chdkHigh) being bigger than the CHDK_DELAY or not. So if (millis() - chdkHigh) wasn't bigger than the delay the first time, the CHDK would never be set back to LOW.

Also, don't return from the function, as there might be other stuff to do, after the CHDK check.