Fix flip-axis math for jobs using a manual start point

[Kreibich04]

isFlipXaxis()/isFlipYaxis() mirror coordinates around bedWidth/bedHeight, which is only valid when the job is referenced to the machine's homed origin. applyStartPoint() re-references job coordinates to a manually jogged start point instead, but the mirror kept using the bed-size pivot, overshooting the flipped axis by roughly bedWidth/bedHeight (X-axis overtravel on machines with the home switch on the flipped side, e.g. OmTech). Pivot is now 0 once a start point is in effect.

Also fixes Ruida's bounding-box (E703/E707/E750-E762) min/max: mirroring around a pivot reverses which bound is min vs max, so each pair needs to be re-sorted after flipping rather than flipped independently.

---

I still need to test this against our laser, if relative positioning is working properly now.

[t-oster]

Hi, why did you close the request?

[Kreibich04]

Hi,
sorry, my comment got lost when closing. When trying it with our laser it did not work as expected and I need to test / troubleshoot further. I plan on raising a PR again, when everything is resolved and ready for review (could have just set it to draft, now that I am thinking about it).

[Kreibich04]

@t-oster created #229 with the adapted & tested changes.