refactor: Fixed the bug to add neck modules properly

[hectorLop]

Fix a bug introduced by the DETR implementation. The problem was that when iterating the named_modules(), it return both the name and the module, but it was iterated as it was a list so the statement if name == 'neck' was never going to be True.

[hectorLop]

Here is a colab notebook, similar to the getting started, but I added a section where I create the YOLOx, Retinanet and DETR models and use fastai to find their optimar learning rates. Thus, it shows that now they can be created succesfully.

[codecov]

Codecov Report

Merging #1029 (43edced) into master (75aac22) will decrease coverage by 0.04%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master    #1029      +/-   ##
==========================================
- Coverage   85.76%   85.71%   -0.05%     
==========================================
  Files         258      258              
  Lines        5676     5679       +3     
==========================================
  Hits         4868     4868              
- Misses        808      811       +3     

Flag	Coverage Δ
unittests	85.71% <0.00%> (-0.05%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
icevision/models/mmdet/utils.py	63.15% <0.00%> (-3.51%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 75aac22...43edced. Read the comment docs.

[FraPochetti]

Thanks a ton again!