should_get: Use format_json.py instead of format-json.sh

format_json.py raises an exception if the JSON is not valid.
Using format_json.py needs to slightly rewrite the test
(spaces are deleted, and the order may not be the same).

algo/tests/stanford-data.should_get

-!LAUNCH: ../../vidjil -G ../../germline/IGH -r 5 -b data  ../../data/Stanford_S22.fasta ; cat out/data.vidjil | sh format-json.sh
+!LAUNCH: ../../vidjil -G ../../germline/IGH -r 5 -b data  ../../data/Stanford_S22.fasta ; cat out/data.vidjil | python ../../tools/format_json.py
 
 $ Number of reads
-1:"total" : [ 13153 ] ,
+e1:"total": [13153],
 
 $ Number of segmented reads
-1:"segmented" :  [ 13139 ] 
+e1:"segmented": [13139]
 
 $ Most abundant window
-1:"id" : "CCACCTATTACTGTACCCGGGAGGAACAATATAGCAGCTGGTACTTTGACTTCTGGGGCC", "reads" : [ 8 ]
+e1:"id": "CCACCTATTACTGTACCCGGGAGGAACAATATAGCAGCTGGTACTTTGACTTCTGGGGCC", "reads": [8]

algo/tests/stanford-fuse.should_get

-!LAUNCH: ../../vidjil -G ../../germline/IGH ../../data/Stanford_S22.fasta ; python ../../tools/fuse.py out/Stanford_S22.vidjil out/Stanford_S22.vidjil -o out/fused.data ; cat out/fused.data | sh format-json.sh
+!LAUNCH: ../../vidjil -G ../../germline/IGH ../../data/Stanford_S22.fasta ; python ../../tools/fuse.py out/Stanford_S22.vidjil out/Stanford_S22.vidjil -o out/fused.data ; cat out/fused.data | python ../../tools/format_json.py
 
 $ Points list
 1:"point": [ "", "" ]
 
 $ Most abundant window, twice, fused 
-1:"id": "CCACCTATTACTGTACCCGGGAGGAACAATATAGCAGCTGGTACTTTGACTTCTGGGGCC", .*"reads": [ 8, 8 ]
+1:"reads": \\[8, 8\\].*"id": "CCACCTATTACTGTACCCGGGAGGAACAATATAGCAGCTGGTACTTTGACTTCTGGGGCC"},.*"top": 2
 
 # Fails since 49046ca6b97, no more 'others'
 $ Windows that are not in the top 50