options: set minimum for "threads" to zero

A negative number of threads does not make sense and 0 is used for
autodetection. Adds a symbolic name for autodetection.

libavcodec/options.c

@@ -370,7 +370,8 @@ static const AVOption options[]={
 {"float", NULL, 0, AV_OPT_TYPE_CONST, {.dbl = FF_AA_FLOAT }, INT_MIN, INT_MAX, V|D, "aa"},
 #endif
 {"qns", "quantizer noise shaping", OFFSET(quantizer_noise_shaping), AV_OPT_TYPE_INT, {.dbl = DEFAULT }, INT_MIN, INT_MAX, V|E},
-{"threads", NULL, OFFSET(thread_count), AV_OPT_TYPE_INT, {.dbl = 1 }, INT_MIN, INT_MAX, V|E|D},
+{"threads", NULL, OFFSET(thread_count), AV_OPT_TYPE_INT, {.dbl = 1 }, 0, INT_MAX, V|E|D, "threads"},
+{"auto", "detect a good number of threads", 0, AV_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, V|E|D, "threads"},
 {"me_threshold", "motion estimaton threshold", OFFSET(me_threshold), AV_OPT_TYPE_INT, {.dbl = DEFAULT }, INT_MIN, INT_MAX, V|E},
 {"mb_threshold", "macroblock threshold", OFFSET(mb_threshold), AV_OPT_TYPE_INT, {.dbl = DEFAULT }, INT_MIN, INT_MAX, V|E},
 {"dc", "intra_dc_precision", OFFSET(intra_dc_precision), AV_OPT_TYPE_INT, {.dbl = 0 }, INT_MIN, INT_MAX, V|E},