faster version of per_file channels

[rettigl]

This PR implements an alternative method to generate the per_file columns, which is much more performant. Additionally, the function to add offsets is modified for a moderate speedup (removing the map_partitions call, and rather directly apply to the df columns)

[coveralls]

Pull Request Test Coverage Report for Build 9628505451

Details

• 33 of 39 (84.62%) changed or added relevant lines in 1 file are covered.
• 2 unchanged lines in 1 file lost coverage.
• Overall coverage increased (+0.008%) to 92.003%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
sed/loader/mpes/loader.py	33	39	84.62%
<!--	Total:	33	39	84.62%	-->

Files with Coverage Reduction	New Missed Lines	%
sed/loader/mpes/loader.py	2	86.81%
<!--	Total:	2		-->

Totals
Change from base Build 9588080384:	0.008%
Covered Lines:	6500
Relevant Lines:	7065

---

💛 - Coveralls

[coveralls]

Pull Request Test Coverage Report for Build 9628850484

Details

• 42 of 48 (87.5%) changed or added relevant lines in 2 files are covered.
• 2 unchanged lines in 1 file lost coverage.
• Overall coverage increased (+0.003%) to 91.998%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
sed/loader/mpes/loader.py	33	39	84.62%
<!--	Total:	42	48	87.5%	-->

Files with Coverage Reduction	New Missed Lines	%
sed/loader/mpes/loader.py	2	86.81%
<!--	Total:	2		-->

Totals
Change from base Build 9588080384:	0.003%
Covered Lines:	6496
Relevant Lines:	7061

---

💛 - Coveralls

[rettigl]

@zain-sohail This solution here passes the benchmarks, even with slight improvements on speed (maybe due to different worker specs), whereas the earlier code consistently is a factor 2-3 slower. I merge this here with the other branch, and then have you review it, if you want to see the changes, you can refer to this PR.