Update AMT / HAMT libraries to allow for ranged iteration

[alexytsu]

Currently the only way to iterate through the contents of a HAMT / AMT is via a for_each that explores every node in the tree.

This is prohibitively expensive in gas costs and disallows paginated APIs to be developed on top of these data structures. In particular the motivating use case of enumerating tokens, owners and balances in the FRC53 NFT standard is not feasible without a paginated API. Therefore, enumeration needs to be scoped to a certain predictable set of ranges.

For both the HAMT and AMT the main gas cost comes from expansion of CID -> Nodes during tree traversal. A limit on the number of blocks expanded/explored will put a limit on gas costs.

By specifying a start range and end range for indexes when iterating through AMTs it is possible to limit the number of nodes explored and thus limit the gas costs.

[alexytsu]

POC implementations are found: https://github.com/helix-onchain/filecoin/pull/185 https://github.com/helix-onchain/ref-fvm/pull/1

[alexytsu]

unresolved API question/design decision: should the range of iteration be specified by:

• index which is potentially a less specific estimation of gas costs but more aligned with high level use cases
• blocks expanded/read which potentially corresponds closer to incurred gas costs but leads to less predictable ranges being specified

[anorth]

I propose letting the caller decide, by providing both values to a predicate that indicates whether to keep gooing

[alexytsu]

Closed along with https://github.com/filecoin-project/ref-fvm/issues/1514