Add our own versions cache

[fthomas]

The same rationale as for #1218 applies here too. Our own versions cache
allows for a more fine-grained control of which calls to Coursier (and
therefore potentially to Maven Central) are rate limited. Since we are
using the same TTL for our cache and for Coursier's cache, our cache
should be as fresh as Coursier's.

The previous rate limiting could result in limited calls that hit
Coursier's cache and unlimited calls that missed Coursier's cache.
This won't happen with this implementation. So we're faster now and
reduce the chance of being blacklisted by Maven Central again.

[codecov]

Codecov Report

Merging #1220 into master will increase coverage by 0.19%.
The diff coverage is 78.57%.

@@            Coverage Diff             @@
##           master    #1220      +/-   ##
==========================================
+ Coverage   68.84%   69.03%   +0.19%     
==========================================
  Files          96       97       +1     
  Lines        1438     1447       +9     
  Branches       35       39       +4     
==========================================
+ Hits          990      999       +9     
  Misses        448      448

Impacted Files	Coverage Δ
.../scalasteward/core/persistence/KeyValueStore.scala	100% <ø> (ø)	⬆️
...a/org/scalasteward/core/coursier/CoursierAlg.scala	100% <ø> (ø)	⬆️
...la/org/scalasteward/core/application/Context.scala	0% <0%> (ø)	⬆️
...scala/org/scalasteward/core/util/RateLimiter.scala	0% <0%> (ø)	⬆️
...scala/org/scalasteward/core/update/UpdateAlg.scala	60% <100%> (-1.91%)	⬇️
...lasteward/core/persistence/JsonKeyValueStore.scala	100% <100%> (+10%)	⬆️
...ain/scala/org/scalasteward/core/data/Version.scala	94.36% <100%> (+0.08%)	⬆️
...rg/scalasteward/core/update/VersionsCacheAlg.scala	85.71% <85.71%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e8e81c5...f995d52. Read the comment docs.