Fix ITS/MFT clusterization for some complex shapes, O2-6424.

[shahor02]

Alternative to #14785

@f3sch, this version avoids looping over preclusters during data processing, just 1 extra linear loop in the finishChip. Does not show any overhead. For you digits created with the filter.C of https://its.cern.ch/jira/browse/O2-6424 I get:

cluster 0 0 0 | BB Top/Left : col 0 row 9 pattID: 25
rowSpan: 5 columnSpan: 3 #bytes: 2
|++ |
|  +|
| ++|
| ++|
|++ |

cluster 0 0 1 | BB Top/Left : col 4 row 4 pattID: 0
rowSpan: 2 columnSpan: 2 #bytes: 1
|++|
|++|

cluster 0 0 2 | BB Top/Left : col 0 row 7 pattID: 1
rowSpan: 1 columnSpan: 1 #bytes: 1
|+|

cluster 0 0 3 | BB Top/Left : col 0 row 18 pattID: 2
rowSpan: 1 columnSpan: 2 #bytes: 1
|++|

cluster 0 0 4 | BB Top/Left : col 0 row 16 pattID: 1
rowSpan: 1 columnSpan: 1 #bytes: 1
|+|

cluster 0 0 5 | BB Top/Left : col 0 row 15 pattID: 1
rowSpan: 1 columnSpan: 1 #bytes: 1
|+|

cluster 0 0 6 | BB Top/Left : col 7 row 15 pattID: 48
rowSpan: 2 columnSpan: 2 #bytes: 1
| +|
|+ |

[github-actions]

REQUEST FOR PRODUCTION RELEASES:
To request your PR to be included in production software, please add the corresponding labels called "async-" to your PR. Add the labels directly (if you have the permissions) or add a comment of the form (note that labels are separated by a ",")

+async-label <label1>, <label2>, !<label3> ...

This will add <label1> and <label2> and removes <label3>.

The following labels are available
async-2023-pbpb-apass4
async-2023-pp-apass4
async-2024-pp-apass1
async-2022-pp-apass7
async-2024-pp-cpass0
async-2024-PbPb-apass1
async-2024-ppRef-apass1
async-2024-PbPb-apass2
async-2023-PbPb-apass5

[f3sch]

@shahor02 very nice!
Just wondering why:

cluster 0 0 4 | BB Top/Left : col 0 row 16 pattID: 1
rowSpan: 1 columnSpan: 1 #bytes: 1
|+|

cluster 0 0 5 | BB Top/Left : col 0 row 15 pattID: 1
rowSpan: 1 columnSpan: 1 #bytes: 1
|+|

are not connected anymore? From the filter.C macro they should be.

[shahor02]

@f3sch, The algorithm assumes that the input digits are ordered in column (inner index) and row (outer). If I swap the order in your macro to

  dv->emplace_back(123, 15, 0, 5);
  dv->emplace_back(123, 16, 0, 5);

then the clusters are:

cluster 0 0 0 | BB Top/Left : col 0 row 9 pattID: 25
rowSpan: 5 columnSpan: 3 #bytes: 2
|++ |
|  +|
| ++|
| ++|
|++ |



cluster 0 0 1 | BB Top/Left : col 4 row 4 pattID: 0
rowSpan: 2 columnSpan: 2 #bytes: 1
|++|
|++|



cluster 0 0 2 | BB Top/Left : col 0 row 7 pattID: 1
rowSpan: 1 columnSpan: 1 #bytes: 1
|+|



cluster 0 0 3 | BB Top/Left : col 0 row 18 pattID: 2
rowSpan: 1 columnSpan: 2 #bytes: 1
|++|



cluster 0 0 4 | BB Top/Left : col 0 row 15 pattID: 3
rowSpan: 2 columnSpan: 1 #bytes: 1
|+|
|+|

[f3sch]

@shahor02 ah yes, than I will close the other PR, thanks for looking into it 👍