Add counts to annotated XML generation

[apriltuesday]

Count and compute simple metrics for comparing CMAT annotations with those in ClinVar. Primary functionality changes are:

• map_genes.py and map_xrefs.py for mapping gene symbols to Ensembl gene ID (using Biomart) and xrefs to EFO synonyms (using OLS)
• annotated_clinvar.py and set_metrics.py for actually computing the metrics
• pipeline.nf for hooking everything together via an optional --evaluate parameter

(Changes in create_efo_table.py and repeat_variant.py are not related, just getting some tests to be less flaky.)

Sample output below:

Overall counts (RCVs):
total                 2352
has_supported_measure 2350
has_supported_trait   1444

Gene annotations:
Total = 2350
        Category  Count  Percent  F1 Score 
     exact_match   2041    86.9%      1.00 
   cmat_superset     33     1.4%      0.67 
     cmat_subset    189     8.0%      0.62 
 divergent_match      1     0.0%      0.50 
        mismatch      1     0.0%      0.00 
 => both_present   2265    96.4%      0.96 
      cv_missing      0     0.0%      0.00 
    cmat_missing     72     3.1%      0.00 
    both_missing     13     0.6%      0.00 

Functional consequences:
Total = 2350
        Category  Count  Percent  F1 Score 
     exact_match   1657    70.5%      1.00 
   cmat_superset      0     0.0%      0.00 
     cmat_subset    608    25.9%      0.64 
 divergent_match      0     0.0%      0.00 
        mismatch      0     0.0%      0.00 
 => both_present   2265    96.4%      0.90 
      cv_missing      0     0.0%      0.00 
    cmat_missing      1     0.0%      0.00 
    both_missing     84     3.6%      0.00 

Trait mappings:
Total = 1686
        Category  Count  Percent  F1 Score 
     exact_match    838    49.7%      1.00 
   cmat_superset     34     2.0%      0.74 
     cmat_subset    420    24.9%      0.66 
 divergent_match     17     1.0%      0.49 
        mismatch    187    11.1%      0.00 
 => both_present   1496    88.7%      0.77 
      cv_missing    162     9.6%      0.00 
    cmat_missing     10     0.6%      0.00 
    both_missing     18     1.1%      0.00 

[apriltuesday]

@tcezard @M-casado I've removed the "work in progress" sample output to avoid confusion, the top-level comment has the most up-to-date info. I realise it's a big PR so feel free to focus your review on the most relevant bits.

[tcezard]

Looks good. A tiny comment.

[tcezard]

Looks good. A tiny comment.

[apriltuesday]

Merging this to fix the memory issue in ClinVarDataset for evidence string generation, as always comments still welcome and will be addressed in a subsequent PR refining the metrics.