SVC client fixup error handling

[mdavidsaver]

In trying out openFPGALoader with a new (and therefore buggy) XVC server I managed to put openFPGALoader into a tight loop printing Send instruction failed when that server crashed.

Somewhat crudely, this PR throws a c++ exception in the places where ll_write() errors were being ignored. These exceptions are not caught, and so abort(). An improvement, although not ideal. It isn't clear to me how to communicate failure back through Xilinx::program().

Also, add sendall() logic as TCP socket send() may enqueue some, but not all, of the requested bytes.

[mdavidsaver]

This is the loop which sticks when shiftIR() returns immediately with an error.

https://github.com/trabucayre/openFPGALoader/blob/39be00fd56e8f532114e86a87f5b1d613159f8dd/src/xilinx.cpp#L727-L729

[trabucayre]

LGTM. Yes usual way is to use return with an error code and let caller deal with this value, but it's true in many part this error code isn't evaluated. Reasons are mainly:

• some cable drivers don't return relevant informations
• coherency issue between them.
• at higher level (vendor level) a correct policy about how to deal with this information must be implemented (stop immediately, recover mode, ...)

[trabucayre]

Applied. Thanks @mdavidsaver