Add warning if there are too many time series

[na--]

Feature Description

Now that k6 v0.41.0 has been released with support for time series, k6 can run out of memory if there are many metric samples with high-cardinality tags. For example, millions of unique URLs (#2762).

Suggested Solution (optional)

RAM usage and availability will vary between instances, but we can choose some fairly high time series count (e.g. 100k, 1m?) that should rarely be reached and emit a warning when it is. In the majority of cases, this will be a problem that the user might not realize.

Ideally, it should be configurable, so the few users that want to intentionally exceed it, can do so without that warning always triggering. However, because of #883, that part might be left for the future. Having the warning sooner rather than later is probably better than a few people seeing an unnecessary warning 🤔

Already existing or connected issues / PRs (optional)

• Significantly high memory usage on 0.41.0? #2762
• Remove the Engine and split metrics/VUs handling #1889
• Refactor the current way metrics are defined #572
• metrics.Ingester: Sink by time series #2735