Remove dependency to SSL in udf client

[tomuben]

Background

Currently, the Java language implementation of the udf client depends on the OpenSSL package (headers+shared libraries). This is, because the ScriptOptions parser needs to calculate MD5 checksums for imported scripts, in order to avoid duplicated imports, see https://github.com/exasol/script-languages/blob/master/exaudfclient/base/javacontainer/script_options/checksum.cc

We can remove the dependency to SSL if we would do a custom implementation of a checksum, which should not be so difficult.

Acceptance Criteria

1. Replace https://github.com/exasol/script-languages/blob/master/exaudfclient/base/javacontainer/script_options/checksum.cc with a custom implementation
2. Add unit tests for the new custom implementation
3. Remove bazel repository SSL

[tomuben]

We decided to use std::unordered_set instead of md5 or any other 3rd party hash algorithm.
The problem with unordered_set is higher amount of memory usage in current implementation.
We propose a two step apprach:

First iteration

Replace checksum calculation with std::unordered_set in parser_ctpg_script_importer.cc and parser_legacy.cc.

This solution will result in higher amount of memory usage in parser_ctpg_script_importer.cc because the scripts will be stored twice in memory:

• In variable collectedScripts
• and in the ordered_set

[tkilias]

After a second discussion, we found that collectdScripts only stores the imported scripts of the current recursion step. So, with time, it accumulates the parts of the whole final script. However, it already contains the imports replaced; for that reason, we can't use it for deduplication.
So, this means we stay with the unordered_set for the moment and take the increased memory usage for more correctness. In practice, huge imports are not to be expected because they are cumbersome. Furthermore, the Java compiler likely uses even more memory.
For a new version of the client, we could think about a deduplication only on the name level instead of the content level. However, this changes the behavior considerably. Furthermore, checking on the name level could also save communication rounds, because for a script we already imported, we don't need to request the script again.

[redcatbear]

The effort went up a lot from our original estimation. 1 pw.