Amend for brotli (de)compress (-mmt1), update README.md

[sebres]

The fix contains:

• if single-threaded brotli, retain artificial number set in BrotliHandler (0) - always prefer it for .br format, otherwise -mmt1 causes that it'd use brotli-mt and the stream is incompatible; -mmt=n with n >= 2 forces multi-threaded compression/decompression (therefore incompatible with standard brotli)
• translate threads to decoder (previously ignored): important for -mmt>=2 to use brotli-mt (the streams are incompatible)
• update README.md - brotli (de)compress, new hash algorithms (amend to #330).

Amend to #351: without the fix, compression with -mmt1 creates brotli-mt stream (incompatible with default brotli) and decompression always used single-threaded compression, so can't unpack the stream, created with -mmt1 at all.

Now the compression and decompression are consistent and:

• without -mmt it is equivalent -mmt0 but also -mmt1 and uses single-threaded brotli for compress/decompress.
• with -mmt=n where n >= 2, it uses multi-threaded brotli (would create or extract incompatible to standard brotli stream).

[sebres]

@mcmilk, by the way, new brotli version (v.1.1.0) got released last weak. I was trying to update it, but Brotli-Adjust.sh seems to be outdated (needs to wrap more headers and renames), no time to fix at the moment. May be instead of such mockups for brotli a static library build as it is (and linking it to 7z) would be better in the long term?

[sebres]

I added to this PR still one simple but important fix - 09a6777b9623618e6c802ae88d1e5ef2938c544d. Without that fix zip type was missing (for instance 7z i | grep zip doesn't contains zip handler anymore), because after adding of brotli code, RegisterArc for zip silently exceeds kNumArcsMax in registering structure g_Arcs.

BTW, I guess this expects some todo later - some error must be thrown for the case we want to register more archive handlers than allowed by kNumArcsMax.