[Idea] Reading contents without verifying structure

Hello Antonio, thanks for the wonderful Bencode library.

Is it achievable to add a feature to read raw-byte contents of a bencode metadata without actually decoding it? For example, I'm sure that this file is a valid .torrent, I checked it's validity once, and I want to read it's contents to get file locations and their sizes.

When we deal with big file list structures, it takes unnecessary resources and waiting time to decode & process them, we can't save the output to cache, since it would double the storage requirements. But reading it as a string is completely other matter.

I got the idea while implementing counting for files, using substr_count() before actually decoding them.

From the tests, it takes ~11 seconds to decode a meta file containing 18k+ files (most modern game releases have ^2 and some ^3 times more) on one core of 1.50Ghz.

Searching for number of strings on a 177k files torrent on the other hand, takes less than a second.