[DPG] tpcSkimsTableCreator: bugfix reserve tables size

[lubynets]

• In the tpcSkimsTableCreator there are two nested loops: collisions and tracks (V0s, cascades). Currently the reservation of the output table size is done inside the collision loop, while it should be done once per dataframe. It brings to underreservation of the size, which had no fatal effect before (although performance could worsen), but after these PRs, DPL Analysis: use unsafe cursor when reserving AliceO2#15450 and AOD Writer: correct reserve for cpvClusters AliceO2#15512 it brings to a fatal error. This PR fixes this logical error by bringing table reservation outside the collision loop;
• Currently the size to be reserved was equal to the number of input tracks (or $2N_{\mathrm{V0}} + N_{\mathrm{casc}}$), which is extremely conservative, since only small part of the input is propagated to the output. This PR introduces configurables, which define the ratio of the expected output table size to that of the input table. The default values are chosen to be by factor of ~5 larger than the "usual" ratio with "common" selection configuration;
• Added QA histograms, where the dataframe-wise distribution of the output/input table size ratios are stored, so one can tune the mentioned configurables if they are too conservative (if they are too aggressive the code will crash).

Tagging @amaringarcia

[github-actions]

O2 linter results: ❌ 1 errors, ⚠️ 0 warnings, 🔕 0 disabled