Correct estimation of overhead of statCache entries

[gargnitingoogle]

Source of discussion: #1514 (comment)

Source code:

gcsfuse/internal/cache/metadata/stat_cache.go

Line 77 in 02b9f18

return uint64(unsafe.Sizeof(gcs.Object{}) + unsafe.Sizeof(entry{}))

This is primarily about the average per-entry size of the statCache structure (

gcsfuse/internal/cache/metadata/stat_cache.go

Line 65 in 02b9f18

type statCache struct {

). Currently I've accounted for the size of the entry structure itself and the size of the gcs.Object structure pointed to by the entry structure.

1. Is sizeof(entry) + sizeof(gcs.Object) a good-enough approximation, or is more needed to be considered for this estimation?
2. The negative cache entries don't actually contain a gcs.Object structure and just have it at nil . So, the Size() function is overestimating for size for such cases. Is there any way to account for it?