rename eos_assert to eosio_assert, implement is_account() api, and move implementation from apply_context.hpp to .cpp

Add block structures to the list of things that may contain ABI-related data in abi_serializer.  fix duplicated native ABI when no wasm API has been published

The wasm spec has a different definition for the call_indirect syntax in textual representation then what we currently support. Add code to support the spec’s syntax as newer binaryen (s2wasm) seems to use the spec’s syntax.

Be aware, this might still not be TECHNICALLY to spec — I think the spec may allow for defining the function prototype right at the call_indirect site. We only support usage of a type name or type index at the call_indirect site.

Previously, a 68GB virtual memory map was created per wasm (contract) instance. This eats up the virtual memory space exceptionally quick.

Now, only allocate 12KB per wasm/contract instance. This is enough for 1024 table elements which is hopefully plenty. The side effect is that every indirect function call now has an additional runtime check on its value (checking if it is less than 1024). This seems like an acceptable tradeoff.

Also I have removed the 4GB alignment for the table and replaced it with 4KB. I wasn’t finding any benefit to that 4GB alignment. The generated x86_64 code was simply in the form of *8(%base, %index) so I probably don’t even need 4KB.

Added checks to all db writes to validate or set the key type and check db reads to validate or ignore if not yet set.

Place some of the wasm_constraints in constexpr variables that can be referred to from elsewhere

the removal of `init` from `setcode` moved compilation AND validation of wasm to the first action.  We should do validation during `setcode` but doing this properly requires additional work.  This is a temporary fix

Additionally, the injector code was failing when hit with hand crafted wasm that had no types/import section(s)

There looks to be an error on how the WAVM code traps table execution on elements beyond maximum table size. It compares the index in to the table to make sure it is less than the table size. But, the table is 16bytes per index -- so the number of bytes in the table is 16 times more than the number of eelements it can hold.

This commit fixes this error. I've confirmed that the LLVM optimizer still removes the entire check -- after all, any UINT32 value can never be greater than or equal to 2^32