Read/Write blockstore should support resuming

[masih]

When writing to a blockstore, the write might fail before the blockstore writes are "finalised/flushed". The API should support resuming writes, picking up where it left off.

This is a must for Ignite requirements.

[willscott]

while each put should ensure the block is durable, we can assume that the need to resume is rare in comparison to the writes, so it is likely an acceptable tradeoff to not serialize the index until finalization, and instead re-generate the partial index of what has been written so far on resumption

[raulk]

For the sake of completeness, another possibility would be to restart the transfer from 0, but this is desirable.

[masih]

Implemented in #147