nit: include but not used

it is used, the DefaultEndpointProvider included this transitively, now it doesnt, so where ever we are using it, we need to to include it now, i.e. same in transfer manager

One thing i wanna point out here since we are wrapping the raw pointer endpointRulesBlob in a ByteCursor (non-owning). The two engines slightly differ in construction:

• RuleEngine's constructor calls aws_endpoints_ruleset_new_from_string which parses the JSON into it's internal DOM (copy)
• BddEngine's constructor calls aws_endpoints_bdd_engine_new_from_bytecode on the partitions blob (copy), but explicitly states that the caller is responsible for keeping the bytecode buffer alive

For the BDD path, should we ensure the lifetime of that cursor by copying it into a Aws::Vector<uint8_t> or Aws::String member member variable before constructing?

ex.

BDDEndpointProvider(const char* endpointRulesBlob, const size_t endpointRulesBlobSz)
      // Copy the bytecode into a member first
      : m_ownedBytecode(reinterpret_cast<const uint8_t*>(endpointRulesBlob),
                        reinterpret_cast<const uint8_t*>(endpointRulesBlob) + endpointRulesBlobSz),
        Base(reinterpret_cast<const char*>(m_ownedBytecode.data()), m_ownedBytecode.size())

good callout, and we solve this through our constexpr ruleset blob, that puts the data firmly in .rodata so we never have to worry about lifetimes. Lets address this the same as CRT and add a loud warning comment. Anyone who will newly subclass this for their own endpoint provider will have to do so explicitly the only path they would get it is if they didnt hav a custom endpoint provider, so it would be a new code path. Lets add a warning to match CRT.