Refinement/Clarification of LongTaskTimer when using Prometheus

[jjank]

Please describe the feature request

Following a question in Slack with @shakuzen https://micrometer-metrics.slack.com/archives/C662HUJC9/p1751460825070189

The Migration to Prometheus 1.x client libs in micrometer 1.13. changed the behaviour for LongTaskTimer.

Specifically in said Slack thread it was concluded that:

• The migration guide could be more precise in outlining two distinct changes
  • Histogram -> GaugeHistogram for LongTaskTimers with histogram buckets.
  • No more active or duration suffix. _active_count -> _count or _gcount, _duration_sum -> _sum or _gsum
• The docs for LongTaskTimer or Micrometer Prometheus docs don't make it clear that .publishPercentileHistogram() might be necessary for certain use cases
• Consider making all LongTaskTimers produce HistogramGauges and ditch Summary due to the semantic mismatch

[shakuzen]

Thanks for opening the issue. If anyone wants to help with a pull request to improve the Micrometer Prometheus docs, it'd be appreciated. I think a table or something showing the mapping from Micrometer types to Prometheus types in different situations would help make things more clear.