Validate current assumption of ledger behavior

[g2flyer]

Description

PR #501 assumes -- as experiments with fabric seemed surprisingly to have shown .. -- that fabric always returns only the commit value, even if the chaincode in the current transaction has issued a putState. On the other hand, fpc currently allows multiple putStates (with the last one counting). As fabric behaviour seems a bit surprising and we might have missed some other subtleties which could lead to behavior which is unexpected to chaincode developer (and potentially insecure) we should cross-check with fabric maintainers ...

Link to feature or bug

Link to dependent issues

[bvavala]

Notes: at the time #501 was merged, FPC updates the read/write set in shim.cpp and behaves as follows for a given key k:

• only the hash of the first read(k) is stored
• the values of subsequent read(k) calls (to Fabric) are checked against the stored hash; if the hash is different, the read fails
• only the value of the most recent write(k,v) is stored