importer schema conversion fix: don't use overly coarse resolution data just because it has a long retention

[replay]

fixes: #1596

[fitzoh]

Would something like this make sense?

func (c *converter) findSmallestLargestArchive(spp, nop uint32) (int, int) {
	smallestArchiveIdx := -1
	largestArchiveIdx := -1
	threshold := nop * spp
	for i, arch := range c.archives {
		// find largest archive that still has a higher or equal resolution than requested
		if arch.SecondsPerPoint <= spp {
			smallestArchiveIdx = i
		}
		// find smallest archive that still contains enough data to satisfy requested range,
		if arch.SecondsPerPoint > threshold {
			break
		}
		largestArchiveIdx = i
	}

	return smallestArchiveIdx, largestArchiveIdx
}

[replay]

thx @fitzoh that looks good.
at some point we should also fix this schema conversion so that it solves this issue as well, but i guess that could be done in a new PR: #1609

[Dieterbe]

I rewrote the history because it was a mess, and came up with a clearer title.