Measures with local time signatures don't work well with the fix for #4867

The fix for #4867: Rests in measures should be grouped according to time signature has been implemented but in cases involving local time signatures it isn't working well. For example let the global time signature be 6/8 and the local time signature be 2/4, which is before the second measure (the problem occurs for other time signatures too but setting them like these makes the problem easy to spot). Add a 16th note to the second measure, and the rests that fill the rest of the measure are not the same as what you'd get if that 2/4 were a global time signature, which is the desired result.

I consider it as a regression because among cases involving local time signatures many more go wrong compared to pre-3.5 versions. Workaround is to manually change the rests.