preview: one preamble for every GUI, restating modes only

[grandixximo]

The preview starts from the machine's state, restated as G-code. AXIS and the gremlin GUIs (gmoccapy, gscreen, qtvcp) each built that preamble their own way, with a bug each:

• AXIS restated stat.mcodes[3], the last block's tool change. After an MDI M61 Q1 a reload ran a bare M61 and failed with "Need non-negative Q-word to specify tool number with M61". Found by Sigma1912.
• gremlin restated a bare G43, which in the preview takes the loaded tool's table offset. After G43 H<other>, G43.1 or G43.2 the preview starts at the wrong height, so moves relative to the start point are drawn off.

preview_helpers.preview_initcodes() is now the one preamble for all of them. It restates only the slots that hold a mode, and the tool length offset in force as G43.1 or G49. It replaces create_unitcode_and_initcode().

I checked it against the tool tip sampled on a sim run: the preview from the shared preamble matches within 0.005 mm in every case, where the gremlin one was 8 to 18 mm off with G43 H<other>, G43.1 and G43.2. tests/interp_initcode now covers G43.1 and a final M61.

[Sigma1912]

Also fixes a bug in gremlin where a gcode (re)loaded against an active G91 would cause an error Cannot use g53 incremental

[grandixximo]

Rebased on master after #4600. Two things changed with it:

• preview_initcodes() now uses axis_kinds.wrapped() for the wrapped rotaries, and restates the tilted work plane with workplane_code() right after the machine units, as AXIS did on master. tests/workplane-preview calls preview_initcodes() now that create_unitcode_and_initcode() is gone.
• The first version left out stat.gcodes[5], the program's G20/G21, on the assumption that the units were already restated. They were not: stat.linear_units is the machine's units, so after an MDI G21 on an inch machine the preview ran the file in inches. tests/workplane-preview caught it once rebased. Slot 5 is restated again, as it was before.