IDRS: Move preconditioning

[h-larsson]

The location of the preconditioner did not seem right. According to [1], it should be applied after v = r - G c.
Currently, the preconditioner is only applied to G c

[1] ACM Trans. Math. Software,, Vol. 38, No. 1, pp. 5:1-5:19, 2011, Fig 23

[coveralls]

Pull Request Test Coverage Report for Build 1383627942

• 0 of 1 (100.0%) changed or added relevant line in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-1.3%) to 96.234%

---

Totals
Change from base Build 888541031:	-1.3%
Covered Lines:	1712
Relevant Lines:	1779

---

💛 - Coveralls

[coveralls]

Pull Request Test Coverage Report for Build 1386804732

• 1 of 1 (100.0%) changed or added relevant line in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 97.525%

---

Totals
Change from base Build 888541031:	0.0%
Covered Lines:	1734
Relevant Lines:	1778

---

💛 - Coveralls

[mschauer]

Thanks for catching that. Would be nice if we can add a test showing that the preconditioner works now

[mschauer]

We could check how the change affects https://github.com/JuliaLinearAlgebra/IterativeSolvers.jl/blob/master/test/idrs.jl#L60

[mschauer]

h-larsson#1 (please merge into your branch) sharpens the tests, in a way that they will fail with the wrong order and pass with your right order. CC @deltaeecs