Fixes race in getKeyMetadata accessor, when called from KeyMetadata

[cognusion]

All(?) cases where cacheShard.getKeyMetadata() was called were wrapped in a lock, except when KeyMetadata was used, resulting in a race between when a key was deleted via cacheShard.removeOldestEntry and read via cacheShard.getKeyMetadata().

WARNING: DATA RACE
Write at 0x00c01697d9b0 by goroutine 133:
  runtime.mapdelete_fast64()
      /home/m/Documents/workspace/go/src/runtime/map_fast64.go:272 +0x0
  github.com/allegro/bigcache/v2.(*cacheShard).removeOldestEntry()
      /home/m/gocode/src/github.com/cognusion/bigcache/shard.go:304 +0x360
  github.com/allegro/bigcache/v2.(*cacheShard).removeOldestEntry-fm()
      /home/m/gocode/src/github.com/cognusion/bigcache/shard.go:293 +0x49
  github.com/allegro/bigcache/v2.(*cacheShard).onEvict()
      /home/m/gocode/src/github.com/cognusion/bigcache/shard.go:249 +0x1cf
  github.com/allegro/bigcache/v2.(*cacheShard).cleanUp()
      /home/m/gocode/src/github.com/cognusion/bigcache/shard.go:260 +0xd8
  github.com/allegro/bigcache/v2.(*BigCache).cleanUp()
      /home/m/gocode/src/github.com/cognusion/bigcache/bigcache.go:215 +0x8b
  github.com/allegro/bigcache/v2.newBigCache.func1()
      /home/m/gocode/src/github.com/cognusion/bigcache/bigcache.go:92 +0xb4

Previous read at 0x00c01697d9b0 by goroutine 28:
  runtime.mapaccess1_fast64()
      /home/m/Documents/workspace/go/src/runtime/map_fast64.go:12 +0x0
  github.com/allegro/bigcache/v2.(*cacheShard).getKeyMetadata()
      /home/m/gocode/src/github.com/cognusion/bigcache/shard.go:348 +0x130
  github.com/allegro/bigcache/v2.(*BigCache).KeyMetadata()
      /home/m/gocode/src/github.com/cognusion/bigcache/bigcache.go:196 +0x14f
  github.com/allegro/bigcache/v2.TestWriteAndReadParallelSameKeyWithStats()
      /home/m/gocode/src/github.com/cognusion/bigcache/bigcache_test.go:699 +0x5f9
  testing.tRunner()
      /home/m/Documents/workspace/go/src/testing/testing.go:991 +0x1eb

[cristaloleg]

Thank you