Fix a valgrind UMR concerning optimised line rendering

[nigels-com]

valgrind warning is seen for first call to render_line
I didn't quite pinpoint the root cause.
memset resolves the warning, possibly faster than the loop?

Before:

==27458== Use of uninitialised value of size 8
==27458==    at 0x1133CB: render_line (video.c:761)
==27458==    by 0x123EA7: video_step (video.c:797)
==27458==    by 0x128695: emulator_loop (main.c:919)
==27458==    by 0x10EC31: main (main.c:788)

After:

==27041== Memcheck, a memory error detector
==27041== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==27041== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==27041== Command: ./x16emu -rom rom.bin -scale 2 -prg petdrawx16.prg
==27041== 
==27041== 
==27041== HEAP SUMMARY:
==27041==     in use at exit: 631,731 bytes in 560 blocks
==27041==   total heap usage: 40,095 allocs, 39,535 frees, 455,145,124 bytes allocated
==27041== 
==27041== LEAK SUMMARY:
==27041==    definitely lost: 8,376 bytes in 6 blocks
==27041==    indirectly lost: 1,728 bytes in 38 blocks
==27041==      possibly lost: 0 bytes in 0 blocks
==27041==    still reachable: 621,627 bytes in 516 blocks
==27041==         suppressed: 0 bytes in 0 blocks
==27041== Rerun with --leak-check=full to see details of leaked memory
==27041== 
==27041== For counts of detected and suppressed errors, rerun with: -v
==27041== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 5 from 1)
``

[nigels-com]

@neilogd I noted your optimisation effort in the recent history, possibly the logic here is clear to you and the specific corner case might be obvious!