Allocation/deallocation bugs

In traipsing through the code, I found a few more bugs. This list includes the `delete _get_pub` bug I reported earlier

```C++
AdafruitIO_Feed.cpp: delete _get_pub; // jmn-bug
AdafruitIO_Group.cpp: // There is no test to see of _sub and _pub allocations succeeded // jmn-bug
AdafruitIO_Group.cpp: // If _sub is NULL, the behavior of the remaining method calls // jmn-bug
AdafruitIO_Group.cpp: // is undefined // jmn-bug
AdafruitIO_Group.cpp: // This code is incorrect; if, for example, _topic was allocated // jmn-bug
AdafruitIO_Group.cpp: // but _create_url failed, then setting topic to NULL (neé 0) // jmn-bug
AdafruitIO_Group.cpp: // will cause a memory leak // jmn-bug
AdafruitIO_Group.cpp: // correct code would be // jmn-bug
AdafruitIO_Group.cpp: // if(_topic != NULL) // jmn-bug
AdafruitIO_Group.cpp: // { /* has topic */ // jmn-bug
AdafruitIO_Group.cpp: // free(_topic); // jmn-bug
AdafruitIO_Group.cpp: // _topic = NULL; // jmn-bug
AdafruitIO_Group.cpp: // } /* has topic */ // jmn-bug
AdafruitIO_Group.cpp: // Repeat for _create_url and _group_url // jmn-bug
```

Start by reviewing the allocation and cleanup paths in AdafruitIO_Feed.cpp and AdafruitIO_Group.cpp, focusing on the lines marked jmn-bug. Trace the listed pointers through failure and deallocation cases. Done means the reported invalid deletion, unchecked allocations, undefined calls, and possible leaks are addressed without introducing new allocation errors.

Written by the indexing model from the issue text.