Fix EcoBraille and Lilly braille display drivers

[LeonarddeR]

Link to issue number:

Fixes #17502

Summary of the issue:

In #15386, support was added for numRows and numCols on braille displays. Setting numCells explicitly would also set numCols.
In #17011, we're much more relying on numCols and numRows.

For the eco braille and Lilli drivers, numCells was implemented by overriding _get_numCells. However this way, the setter of numCells would never be called and therefore numCols would always be 0.

Description of user facing changes

Fix braille for Eco Braille and Lilli displays.

Description of development approach

1. When calculating display dimensions on the braille handler, when display.numRows is 1, use display.numCells instead of display.numCols. This accounts for cases where numCells is implemented and numRows/numCols are not.
2. Worked around the ZeroDevisionError when getting windowStartPost on a buffer
3. Initialize _windowRowBufferOffsets to [(0, 0)] instead of [(0, 1)]. The latter suggests that the window contains one cell, which is not the case. @michaelDCurran you might want to check this.

Testing strategy:

@dreinn Could you test this?

Known issues with pull request:

I think we should consider a long term strategy for numCells vs. numRows/numCols. We still need to account for both now, which isn't ideal and could easily go wrong as proved.

Code Review Checklist:

• Documentation:
  • Change log entry
  • User Documentation
  • Developer / Technical Documentation
  • Context sensitive help for GUI changes
• Testing:
  • Unit tests
  • System (end to end) tests
  • Manual testing
• UX of all users considered:
  • Speech
  • Braille
  • Low Vision
  • Different web browsers
  • Localization in other languages / culture than English
• API is compatible with existing add-ons.
• Security precautions taken.

@coderabbitai summary

[AppVeyorBot]

• PASS: Translation comments check.
• PASS: License check.
• PASS: Unit tests.
• PASS: Lint check.
• FAIL: System tests (tags: installer NVDA). See test results for more information.
• Build (for testing PR): https://ci.appveyor.com/api/buildjobs/vd27loxwdp04kklc/artifacts/output/nvda_snapshot_pr17635-35058,a1fe5402.exe
• CI timing (mins):
  INIT 0.0,
  INSTALL_START 1.3,
  INSTALL_END 1.0,
  BUILD_START 0.0,
  BUILD_END 27.6,
  TESTSETUP_START 0.0,
  TESTSETUP_END 0.4,
  TEST_START 0.0,
  TEST_END 18.3,
  FINISH_END 0.2

See test results for failed build of commit a1fe5402ce

[SaschaCowley]

Thanks for this, @LeonarddeR. Just a docstring change while you're here. Plus we need to wait on real world testing.

[LeonarddeR]

I asked an additional review from @michaelDCurran since this touches his code. I'd like to make sure that he agrees with changing the initial state of _windowRowBufferOffsets to reflect that there's no window.