New implementation for sifting

[henry2004y]

This is a PR for exploring the possibility of shifting towards the code structure of cg.jl in IterativeSolvers.jl. Compared with the current implementation, it shares the same ideas but in my view easier to follow and maintain:

• Now there is only one struct SiftIterable for storing sifting variables, instead of one SiftIterable and one SiftState. All the input variables and intermediate variables are kept in the same place, which may make it easier to add other extensions.
• There is only one iterate method for SiftIterable, and another sift_iterator method for initializing SiftIterable. I think this is easier to understand than two dispatched iterate methods.

Besides, I changed some parts of the variable names, e.g. maxes --> maxs, yvec --> y, but we can always make better names later.

Take a look and we can discuss more about this!

[henry2004y]

Probably I should rebase instead of merge... Forgive me as a terrible git user

[coveralls]

Pull Request Test Coverage Report for Build 52

• 37 of 42 (88.1%) changed or added relevant lines in 2 files are covered.
• 10 unchanged lines in 2 files lost coverage.
• Overall coverage increased (+3.4%) to 50.811%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/EmpiricalModeDecomposition.jl	1	6	16.67%

Files with Coverage Reduction	New Missed Lines	%
src/testdata.jl	4	0%
src/EmpiricalModeDecomposition.jl	6	34.21%

Totals
Change from base Build 49:	3.4%
Covered Lines:	94
Relevant Lines:	185

---

💛 - Coveralls

[henry2004y]

This new implementation actually does not give identical results as the original version. I need to figure out why.

The reason is that I wrote imf=y before creating SiftIterable. I should either say imf = copy(y), or imf .= y. Fixed in 2d25f77