Detect SPDX Snippets

[mxmehl]

Fixes #696

This PR shall consist of two steps:

1. Allow that the full file is analysed for license/copyright info if we have indicators that they contain a SPDX snippet
2. Actually understand the licensing/copyright info in a snippet (specifically the SPDX-SnippetCopyrightText tag)

On 1: Based on the results of the performance comparison, I implemented a logic that first scans the file for a string (as bytes) (SPDX-SnippetBegin), and resets the reading buffer. If a match is found, the _HEADER_SIZE limit (currently 4096 bytes) is unset so the full file is fed into extract_spdx_info(decoded_text_from_binary()).

Amazingly, the performance ramifications are minimal! On multiple runs against the Linux kernel, the negative effect is ~0.1s or ~1.3%). This is what I would consider an acceptable performance loss, especially in comparison with the 140% loss we'd have if we just removed the limit.

On 2: Treat SPDX-SnippetCopyrightText the same as other (file-level) copyright tags. Of course it's not the same but for REUSE's purpose it should suffice.

However, we could check whether a snippet has licensing and copyright info. That may be a lot trickier as we would have to understand the boundaries of a snippet, and treat it isolatedly. I'm not sure whether it's worth to invest time into this now, or later. Interested in others' opinions.

[linozen]

LGTM 👍

[mxmehl]

However, we could check whether a snippet has licensing and copyright info. That may be a lot trickier as we would have to understand the boundaries of a snippet, and treat it isolatedly. I'm not sure whether it's worth to invest time into this now, or later. Interested in others' opinions.

Linus, Nico and I agreed that, while this would be interesting information, we don't plan to integrate this as of now. It could be a future improvement.