Sanity/bug check for safety scan failing during encode

Ran some tests on how common this failure mode is, and 1-2% of frames
tend to be "bad" (and thus are thrown out). We should never ever hit
this 5-in-a-row case, but it'd also be pretty brutal to infinite loop
due to a bug in the will_it_scan() function...

So, paranoia for now

src/lib/encoder/Encoder.h

@@ -56,6 +56,7 @@ inline unsigned Encoder::encode_fountain(const std::string& filename, const std:
 		requiredFrames = 1;
 
 	unsigned i = 0;
+	unsigned consecutiveScansFailed = 0;
 	while (i < requiredFrames)
 	{
 		auto frame = encode_next(*fes, canvas_size);
@@ -67,8 +68,15 @@ inline unsigned Encoder::encode_fountain(const std::string& filename, const std:
 		// corner "anchors" and fail to extract. So:
 		// if frame fails the scan, skip it.
 		if (!Scanner::will_it_scan(*frame))
-			continue;
+		{
+			if (++consecutiveScansFailed < 5)
+				continue;
 
+			// else, we gotta make forward progress. And it's probably a bug?
+			std::cerr << fmt::format("generated {} bad frames in a row. This really shouldn't happen, maybe report a bug. :(", consecutiveScansFailed) << std::endl;
+		}
+
+		consecutiveScansFailed = 0;
 		if (!on_frame(*frame, i))
 			break;
 		++i;