Implement mass-concentrations from Ishiyama et al. (c500)

[nikfilippas]

Addresses #892 .

Checklist:

• mass-concentration relation
• M500 mass definition
• unit tests
• benchmarks

[coveralls]

Pull Request Test Coverage Report for Build 1689806813

• 127 of 128 (99.22%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.2%) to 97.047%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
pyccl/halos/concentration.py	124	125	99.2%

Totals
Change from base Build 1532031833:	0.2%
Covered Lines:	4403
Relevant Lines:	4537

---

💛 - Coveralls

[c-d-leonard]

Is this ready for review? @nikfilippas

[nikfilippas]

@c-d-leonard not yet! The science is all there but the computation is very slow. Once I find some time I want to speed it up by ~1 order of magnitude using a shortcut in CCL. I'll ask for someone to review this when it's all done.

[damonge]

Good stuff @nikfilippas !
See below a few minor comments. I'll check timing on my end while you address these

[damonge]

Just a quick note: I just checked and I don't think the computational bottleneck comes from the calculation of dlnsigma/dlnR. If anything it may be dominated by the root finding, but that it's quite fast in itself, so I don't think we should worry.
In summary: ready to go as soon as the previous comments are addressed.

[nikfilippas]

You are right, the bottleneck is the computation of sigmaM, in case these haven't already been computed. Then, root finding also takes time but I can't think of any other way. I profiled this one and Duffy08, and this one is an order of magnitude slower. Have a look at the attached images.
Ishiyama21

Duffy08

[nikfilippas]

@damonge ready to be looked at again.

[damonge]

OK, you'll have to teach me how to generate those timing diagrams ;-)

Just one comment above, and we can merge as soon as that's done.

[damonge]

LGTM