fix: a set of incorrect behaviors

[usamoi]

• set amoptionalkey to false: it forbids PostgreSQL to generate a plan to perform a vector index scan without keys or orderByKeys; it does not lead to problems now because we set cost of such a path to an extremely large value.
• set returned value of aminsert to false: it's not significant but we'd better return false
• forbid scanning with a non-MVCC-compliant snapshot: if we get some physical pointers from service and they are just deleted both in a heap table and service, attemptions to get tuples by these physical pointers will fail; if some rows with these physical pointers are inserted, it even produces incorrect results. To solve it, we temporarily forbid scanning with a non-MVCC-compliant snapshot.
• pg_vector_index_stat: we forgot to pass index oid after feat: synchronous remote index dropping #414 is implemented
• printing info instead of throwing errors in amvalidate: it's recommended behavior in PostgreSQL document.