Support uncompressed or fully in-memory vocabularies

[joka921]

So far, QLever's vocabulary (the set of distinct IRIs and literals) is compressed (on the level of individual IRIs and literals) and resides on disk by default. It can be configured that certain IRIs or literals reside in RAM. However, there was some overhead for determining whether an IRI or literal resides in memory or on disk. With this change, a free choice between the following four kinds of vocabularies:

1. On disk and compressed (the default so far)
2. On disk and uncompressed (costs more space but saves time for decompression)
3. Fully in memory and compressed (faster than disk but memory is expensive)
4. Fully in memory and uncompressed (maximally fast but most memory hungry)

Performance tests for exporting a TSV with 100M triples and a total size of 8.8 GB yielded the following result on a Ryzen 9 9950X with 4 x WD_BLACK SN850X NVMe RAID0 (column 2) and a Ryzen 5900X with 1 x TODO NVMe (column 3):

| on disk, compressed | 65s | 177s |
| on disk, uncompressed | 58s | 151s |
| in memory, compressed | 51s | 110s |
| in memory, uncompressed | 50s | 109s |

When measuring the performance multiple times in a row, so that the relevant parts of the disk-based vocabulary are cached by the operating system, the measurements from the first two rows decrease slightly (this effect is more pronounced when exporting smaller results).

BOTTOM LINE: An in-memory vocabulary is 40 - 60% faster for a single SSD, but only 5 - 15% faster for four parallel SSDs. Using compression is around 10% slower on disk and comes at almost no cost in memory.

[codecov]

Codecov Report

Attention: Patch coverage is 99.43182% with 1 line in your changes missing coverage. Please review.

Project coverage is 90.03%. Comparing base (098b79c) to head (d8080b3).
Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/index/Vocabulary.cpp	90.90%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1740      +/-   ##
==========================================
+ Coverage   90.01%   90.03%   +0.01%     
==========================================
  Files         395      398       +3     
  Lines       37793    37954     +161     
  Branches     4257     4262       +5     
==========================================
+ Hits        34019    34171     +152     
- Misses       2478     2482       +4     
- Partials     1296     1301       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[hannahbast]

Quick 1-1 with Johannes, this looks great and it works!

[sparql-conformance]

Conformance check passed ✅

No test result changes.

Details: https://qlever.cs.uni-freiburg.de/sparql-conformance-ui?cur=d8080b30f9914a89e3ed3dcda9c9ccf85a880795&prev=a307781592842f82bdfef78fc47fc832fd37368d

[sonarqubecloud]

Quality Gate failed

Failed conditions
E Reliability Rating on New Code (required ≥ A)

See analysis details on SonarQube Cloud

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE