cache key cant be reused when an interval changes

[owen-d]

What this PR does / why we need it:
This fixes a small (and probably unrealistic) loophole in the cache key generation code for the result cache. Including the interval used in a cache key will prevent us from ever reusing the same key across querier.split-queries-by-interval (and associated overrides) configuration changes.

For a hypothetical request start value 12 and split intervals 5 and 6, we'd generate the same cache keys because 12/5 == 12/6 in integer division. We now include the 5 and 6 respectively in our cache keys to prevent this collision.

Checklist

• Tests updated

[codecov-io]

Codecov Report

Merging #1616 into master will decrease coverage by 0.1%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #1616      +/-   ##
==========================================
- Coverage   61.02%   60.91%   -0.11%     
==========================================
  Files         108      108              
  Lines        8143     8144       +1     
==========================================
- Hits         4969     4961       -8     
- Misses       2786     2791       +5     
- Partials      388      392       +4

Impacted Files	Coverage Δ
pkg/querier/queryrange/limits.go	100% <100%> (ø)	⬆️
pkg/promtail/targets/tailer.go	73.56% <0%> (-4.6%)	⬇️
pkg/promtail/targets/filetarget.go	68.71% <0%> (-1.85%)	⬇️
pkg/logql/evaluator.go	91.75% <0%> (-1.04%)	⬇️

[cyriltovena]

LGTM