fixes #2361: add note about --force when using -u / -n

docs/hashcat-plugin-development-guide.md

@@ -528,7 +528,7 @@ Additionally there is a couple of command line parameters that you want to use:
 
 * --potfile-disable: The moment when your implementation is almost complete and you start cracking the hash for the first time as expected it is very likely it will turn out some other things are not perfect. For instance, the encoding of the cracked hash. So you need to change some code and run hashcat again to verify this option will prevent hashcat from writing the cracked hash to the potfile. This will allow you to restart hashcat without the need to remove the potfile manually.
 * --self-test-disable: The self-test feature serves to test the kernel each time the user starts hashcat to ensure it works on the users hardware as expected. Hashcat does not know you are implementing a new kernel, so it will call the kernel you are implementing, too. This has two unwanted side-effects. First, it will print a self-test failure which is clear to us, but not to hashcat. Second and more relevant, if you use printf(), it is very likely to print values which you are not expecting. This is because you are expecting values based on the hash or password you gave on the command line, not the values produced by the self-test hash or self-test password. If you use the same hash and password, you may wonder why it is printed twice.
-* -n 1 -u 1 -T 1: The combination of these three options with these exact values will disable the auto tune. This is hardcoded into hashcat. This is an undocumented feature. The auto tune will create the same problems as the self-test feature.
+* -n 1 -u 1 -T 1: The combination of these three options with these exact values will disable the auto tune. This is hardcoded into hashcat. This is an undocumented feature. The auto tune will create the same problems as the self-test feature. You need to use --force to be able to use these manual tuning parameters.
 * --quiet: When you are expecting printf() results, try to limit the hashcat output to a minimum. The printf() itself is not affected by this option.
 * --backend-vector-width 1: Only required if you are developing the kernel by using a CPU as compute device. Printing elements from vector data types is possible (for instance `printf (a.s1);`), but we should avoid any influence. Some OpenCL runtimes even support printf() of a vector data type, which results in very weird outputs.
 * -d 1: In case you have multi compute devices in your system, limit it to a single compute device. This is to reduce startup and JiT compile time.