Per-needle x-stitch-number

[ixchow]

Currently, the backend starts a new pass for each new stitch number; but the kniterate does not require this.

Proposed fix:
add a "stitch-numbers" array to each pass along with the slots array and store stitch numbers into it, only breaking passes when conflicting stitch numbers for a given slot are found.

Alternate fix: add a "stitch-number" field to each slot in the slots array.

[ixchow]

Some more questions that came up in implementation:
What stitch number should needles which don't have stitches on them use?
Possible resolutions:

• the number from the closest same-bed needle that has a stitch assigned
• a linear interpolation between the two adjacent stitch numbers (only makes sense if stitch numbers map directly to lengths)
• the number from the next (in pass direction) needle that has a stitch assigned (idea: give stitch cam as much time as possible to switch settings; only makes sense if machine firmware isn't already doing some sort of look-ahead)
• a 'default' or 'empty' value. (is that 0?)

What stitch number should be used on a bed with no stitches on it?
Possible answers:

• The stitch cam values from the other bed (directly copied)
• The most-used value from the other bed
• A 'default' or 'empty' value. (0?)

For a first pass of implementation, I am leaning toward choosing "the next value in pass direction" and "default/empty (0)".

[ixchow]

Preliminary support added in 0d269b2 ; implementation stores stitch values in every slot and uses within-pass lookahead to fill missing stitch values. In case of passes with no needles set to a particular stitch cam value, 0 is used for the whole bed.

It would likely be useful to use further lookahead to resolve the stitch value ambiguity.

Note, I also added x-xfer-stitch-number to set stitch numbers independently for xfer operations.