Fix a regression in decoding of dynamic structs!

[haltman-at]

It seems I accidentally broke decoding of dynamic structs in calldata a while back. Probably when I factored out struct decoding; in combining the static and dynamic cases, I left out the startPosition argument in the call to decodeAbi, which is fine in the static case but is very wrong in the dynamic case. Oops.

(In the static case, you're always reading the member directly, not a pointer to it, so you don't need to know what base that pointer should be considered relative to, because it's not a pointer at all. In the dynamic case, you may be reading a relative pointer, so you need to know what it's relative to.)

As a check against this happening again, I added an additional (very simple) calldata decoding test to catch this.

Note that in order to make this work I ended up updating the Solidity version used in our debugger tests to 0.5.10. (It was on 0.5.4; dynamic structs in calldata weren't allowed yet in 0.5.4.)

[gnidan]

hooray for regression test!

looks good - feel free to take or leave the suggestions

[coveralls]

Coverage remained the same at 70.16% when pulling 1700933 on fix-dynamic-struct-regression into ddc5216 on develop.