Correctly decode UTF-8

[haltman-at]

I originally made this change on type-and-value, then realized it was worth backporting to develop. (I'm setting myself up for a merge conflict, but...)

This PR fixes an issue in string decoding by using the utf8 library to convert Solidity's UTF-8 to Javascript's UTF-16. Note that Solidity does allow malformed UTF-8; in this case, we simply return null. (Note that we detect this case by seeing whether utf8.decode throws an error; it doesn't seem to have a check function separate from the decode.)

Why null rather than undefined? To avoid throwing the debugger into an infinite loop in case a mapping key is a malformed string. Yes, this is a bit of a hack, but don't worry, it'll be gone in 5.1; on type-and-value we return an object representing the particular malformed string rather than a simple null. But here on develop, without the new output format, null will have to suffice.

[coveralls]

Coverage remained the same at 70.16% when pulling cf26256 on utf8-develop into 5071514 on develop.