Guard urclock_getsync() against initial spurious input

[stefanrueger]

Fixes Issue #1245

[stefanrueger]

Under Linux with urclock boards and legacy bootloaders no -xdelay=... is needed. The following bash script tests 20 delays from -15 ms to 80 ms.

for d in {-15..80..5}; do
  avrdude -curclock -pm328p -xdelay=$d -xshowall >& /dev/null
  echo -n $? | tr 01 .X
done; echo

A . means the -xdelay= value works; an X means it fails. Delay 0 (absence of -xdelay=...) is the fourth result letter:

Bootloader	Result	Result with -xstrict	Comments
u7.5 we--h-r-.	....................	....................	All good
u7.6 w-u-jpr-.	....................	....................	All good
u7.7 weu-jPr-c	....................	....................	All good
u7.7 weu-jPrac	X...................	....................	-xdelay=n OK for n >= -10
o8.3 --s-h-r--	....................	X............X......	Some -xdelay do not work
o4.4 --s-h-r--	....................	XXXXXXXXXXXXXXXXXXXX	optiboot v4.4 must not use -xstrict
x0.0 .........	XXXXXXXXXXXXXXXXXXXX	....................	picoboot must use -xstrict

[stefanrueger]

Decided to lengthen delay after issuing reset for benefit of autobaud bootloaders: it really is the first byte from avrdude that makes autobaud bootloaders set their USART divisor, so we better make sure the board has had enough time to come out of reset. @mcuee had a Nano clone board that needed a bit longer (under Windows) so better to increase the margin.

Also changed from warning to notice2 the info on libreadline.

So, am going to merge so we are ready for a New Year release :)