feat(transcripts): Implement source filtering for transcript sampling

[yanirmr]

• Add DESIREDD_SOURCES configuration to prioritize transcript sources
• Modify initialize_transcripts() to create filtered view during init
• Update get_content() to sample only from desired sources

[yoadsn]

@yanirmr Seems to me like this makes the choice exclusive - meaning the sources specified alone would provide transcriptions to users. My expectation that you would increase their likelihood to appear - so we still sample from all the population. So you could configure something like "diversity correction bias ratio" to be 0..1 and if it's 1 all sources are from the preferred and if 0 will be like now across all population.

This makes more sense to me.