Text-to-speech management (fixes #197, #170)

[digital-carver]

• Redirects the STDERR output of where command used in User.has_command? check. We deal with the result inside the program anyway, so error messages like in Error: "which: no say in (...$path...)" #170 and Error: no say in #197 are unnecessary.

• Checks that a known text-to-speech program is available before turning speech ON.

• Checks for command availability before performing various voice-related actions. For eg., currently betty list your voices when run in a system without the say command gives the output:

Betty: Running say -v "?"
sh: say: command not found

whereas with these changes, betty list your voices returns:

Betty: I don't seem to have a voice at all (the program 'say' would help me get some).

• Some minor cleanup of the speak() function

[digital-carver]

The has_command? change also fixes things like #189, which is most likely an incidental error message due to the invocation of the where command.