[io] Validate ZSTD target/source sizes and compr level

[dpiparo]

This PR fixes #9334

[phsft-bot]

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default
How to customize builds

[dpiparo]

I found line https://github.com/root-project/root/pull/15038/files#diff-c35e46fbb246941a30fd3fa261506e5eb03b601e3bf6734460f814fd3aa156c9R46 intriguing. We are doubling the compression level given in input. The ZSTD documentation does not say anything which could possibly justify this behaviour https://facebook.github.io/zstd/zstd_manual.html . This deserves a bit of caution/discussion I believe, also given the test of the compression level which has to be coherent with that.

[github-actions]

Test Results

    12 files      12 suites   1d 23h 50m 36s ⏱️
 2 603 tests  2 603 ✅ 0 💤 0 ❌
29 185 runs  29 185 ✅ 0 💤 0 ❌

Results for commit 5058f06.

[hahnjo]

I found line https://github.com/root-project/root/pull/15038/files#diff-c35e46fbb246941a30fd3fa261506e5eb03b601e3bf6734460f814fd3aa156c9R46 intriguing. We are doubling the compression level given in input. The ZSTD documentation does not say anything which could possibly justify this behaviour https://facebook.github.io/zstd/zstd_manual.html . This deserves a bit of caution/discussion I believe, also given the test of the compression level which has to be coherent with that.

From the linked manual:

The library supports regular compression levels from 1 up to ZSTD_maxCLevel(),
which is currently 22. Levels >= 20, labeled --ultra, should be used with
caution, as they require more memory. The library also offers negative
compression levels, which extend the range of speed vs. ratio preferences.
The lower the level, the faster the speed (at the cost of compression).

So the doubling maps ROOT's single digit layer to the "reasonable" level 0 to 18.

[hahnjo]

As mentioned inline, the check of tgtsize is not needed since 23261a6. The check for *srcsize > 0xffffff should likely be moved to the upper layer with proper error reporting (not just silently returning).