Function `tfw_http_meth_str2id` is totally wrong

[EvgeniiMekhanik]

In tfw_http_meth_str2id function we check first letter in chunk and the length of header and return method type according to it. It is totally wrong, because we return method 'PUT' if the real method was 'PAT' or 'POT'
and so on. This function should be reworked

[krizhanovsky]

Why do we allow unknown methods at first place? Who uses such methods?

I believe we need to block all unknown methods on the HTTP parser and HPACK levels. This will make the parser faster, more robust and secure. See commit 73d30d4

This should be done as the transition to observability #2399

[krizhanovsky]

Unknown methods were introduced in 03aac71 . The test https://github.com/tempesta-tech/tempesta/blob/master/fw/t/unit/test_http1_parser.c#L97 enumerates plenty of methods, unsupported by Tempesta FW, but still common in the Intenet, e.g. CONNECT or VERSION-CONTROL from WebDAV.

It makes sense to preserve the ability to proxy unknown headers, but yes, we should go with method PUT only if we explicitly parsed PUT, not PUTX or soemething like this.