[Win32] Adapt further places to round up when computing control size

[HeikoKlare]

Several places in the computations of preferred sizes of controls used commercial rounding instead of rounding up in pixel/point calculations. This change further adapts additional places to ensure that the calculated size will be large enough and and subsequent point-to-pixel calculation when setting the bounds according the computations does not lead to smaller results.

This is a follow up to:

• [Win32] Pass point values to size computation methods #2717

I did not find any visual differences yet, so consider this a unification/cleanup to ensure that every conversion involved in computeSize() calculations rounds up.

[github-actions]

Test Results

  118 files  + 7    118 suites  +7   17m 41s ⏱️ + 2m 35s
4 651 tests +56  4 634 ✅ +72  17 💤 +3  0 ❌  - 6 
  330 runs  +56    326 ✅ +53   4 💤 +3  0 ❌ ±0 

Results for commit 4a58a39. ± Comparison against base commit 9119357.

♻️ This comment has been updated with latest results.