audioaggregator: Output buffer size should be configurable as fraction

Submitted by Sebastian Dröge @slomo

Link to original bug (#797081)

Description

Similar to how it works in audiobuffersplit, with compensation for rounding errors. This would allow us to output e.g. 30000/1001 audio buffers per second and have them the same size as the video without first going through audiobuffersplit.

Problem is that the property is currently in nanoseconds. We could add a new one and make the old one deprecated, and return rounded nanoseconds from it when reading the old property.

Thoughts?