investigate high resources consumption

[poolpOrg]

It seems that recent changes have improved RAM consumption on initial push...

... but secondary push and some commands, like check, exhaust RAM resources.

This is counter-intuitie since first push is supposedly the most costly operation and check is supposedly very cheap, so something needs to be investigated here.

[poolpOrg]

identified as side effects from ram -> ldb switch and a missing tmp cleanup: leads mfs tmp directories to retain objects in memory.

as an additional improvement I made sure we don't keep too many vfs and indexes in memory simultaneously, the refactor is committed in branch ldb-cleanup pending some testing before merging to main.

[poolpOrg]

Lots of rework was done in the VFS layer, the only remaining part using memory is states which is the topic of issue #90