[bug] `layout.get_collections('run', merge=True)` extra subject `0`; missing others

[WillForan]

Using merge=True with get_collections introduces rows where subject='0'.
Without subject/session filters, get_collections('run',merge=True) also drops whole subjects.
Likely related to #604 (issues with to_df())

merge on just the first subject and session yields rows with subject as 0

layout = bids.BIDSLayout('tests/data/synthetic')
r = runc.variables['respiratory'].to_df()

runc_bug = layout.get_collections("run", subject="01", session="01", merge=True) 
runc_bug.variables['respiratory'].to_df().subject.unique()
# BUG: expect '01' get '01' and 0
# array(['01', 0], dtype=object)

No filter merge also drops subjs 02 to 04

# without subject= and session= filters, missing entire subjects
runc_all = layout.get_collections("run", merge=True)
runc_all.variables['respiratory'].to_df().subject.unique()
# BUG: expect 01-05. missing middle
# array(['01', '05', 0], dtype=object)

Without merge (no internal to_df()?), there's no subject 0 and all subject sessions are in the dat.

# as expected without merge=False
runc_nomerge = layout.get_collections("run", subject="01", session="01")
runc_nomerge[0].variables['respiratory'].to_df().subject.unique()
# array(['01'], dtype=object)

# and without filters
runc_nomerge_all = layout.get_collections("run")
len(runc_nomerge_all) # 30, subj * ses * task
[x.variables['respiratory'].to_df().subject.unique() for x in runc_nomerge_all ]
# [['01'], ['01'], ['01'], ['01'], ['01'], ['01'], ['02'], ['02'], ['02'], ['02'], ['02'], ['02'], ['03'], ['03'], ['03'], ['03'], ['03'], ['03'], ['04'], ['04'], ['04'], ['04'], ['04'], ['04'], ['05'], ['05'], ['05'], ['05'], ['05'], ['05']]

Initial debugging shows rest row (without run in entities) as introducing the subject==0 row

r = runc_bug.variables['respiratory'].to_df()
r.groupby(["subject",  "session",  "task",   "TaskName",   "run",
           "suffix", "EchoTime",  "RepetitionTime",
           "datatype", "extension",  "condition",  "duration", ]).agg(
    {k: (np.mean, np.max, lambda x: len(pd.unique(x))) 
       for k in ["amplitude", "onset"]}
)


                                                                                                          amplitude                        onset                  
                                                                                                               mean       max <lambda_0>    mean    max <lambda_0>
subject session task  TaskName run suffix EchoTime RepetitionTime datatype extension condition   duration                                                         
0       0       0     0        0   0      0.00     0.0            0        0         respiratory 0.1       0.000537  0.999607       1363  399.95  479.9       1600
01      01      nback N-Back   1   bold   0.03     2.5            func     .nii      respiratory 0.1      -0.000794  0.999792       1150   79.95  159.9       1600
                rest  Rest     2   bold   0.03     2.5            func     .nii      respiratory 0.1      -0.014537  0.999752       1145  239.95  319.9       1600

Source of the bug might be that the rest df does not have a run column

dfs = [x.variables['respiratory'].to_df() for x in runc_nomerge]
print("\n".join([str(df.head(n=2)) for df in dfs]))


   amplitude    condition EchoTime RepetitionTime TaskName datatype extension run session subject suffix   task  onset  duration
0        0.0  respiratory     0.03            2.5   N-Back     func      .nii   1      01      01   bold  nback    0.0       0.1
1        0.0  respiratory     0.03            2.5   N-Back     func      .nii   1      01      01   bold  nback    0.1       0.1
   amplitude    condition EchoTime RepetitionTime TaskName datatype extension run session subject suffix   task  onset  duration
0        0.0  respiratory     0.03            2.5   N-Back     func      .nii   2      01      01   bold  nback    0.0       0.1
1        0.0  respiratory     0.03            2.5   N-Back     func      .nii   2      01      01   bold  nback    0.1       0.1
   amplitude    condition EchoTime RepetitionTime TaskName datatype extension session subject suffix  task  onset  duration
0  -0.825171  respiratory     0.03            2.5     Rest     func      .nii      01      01   bold  rest    0.0       0.1
1  -0.788074  respiratory     0.03            2.5     Rest     func      .nii      01      01   bold  rest    0.1       0.1

[WillForan]

I think it's fixed in #1272. there was a missing += when setting row ranges to update the merged dataframe.

test demonstrating also in the PR