bugfix/layer progress mainsail parity

[gmmcosta15]

Description

Select the type:

• Feature
• Bug fix
• Code refactor
• Documentation

Fixes the layer-count and progress-bar behaviour reported from the field. Follow-up to #269 and #272, whose stateless-layer approach did not hold on device.

• Freeze the layer across a pause (set on pause click and paused state, released on printing) so the Happy Hare park Z-lift no longer bumps it to a bogus value.
• Reset layer/total/progress on a new filename, so Mainsail-started prints clear stale values (previously only screen-started prints reset).
• Compute progress as Mainsail file-relative (getPrintPercentByFilepositionRelative): clip file position to gcode_start_byte/gcode_end_byte, so start macros read 0% and the tail matches Mainsail; falls back to raw virtual_sdcard.progress.
• On completion show 100% and the final layer = total, and stop blanking the counters.
• Replace CustomProgressBar.setValue(float) (incompatible override, wrong 0-100 bound) with a clamped set_progress(fraction) and add a real reset() that clears the painted value.

Motivation

Progress used raw absolute file position, so start-print macros (mesh/plate calibration) showed ~4% and the end diverged from Mainsail.The pause layer jump and the stale/blank layer readouts came from the reset living only in the screen-initiated path and from Z-fallback recompute racing the pause state update.

Tests

11 new unit tests in test_job_status_page_unit.py, plus new test_progress_bar_unit.py).