Squash commits to prepare for v2

[jlapeyre]

The main change in the API is that functions in IsApprox with the same name as those in Base and LinearAlgebra are now different methods of the same function. So you can do using of both LinearAlgebra and IsApprox withouth qualifying any symbols.

Another API change making the signatures uniform, rather than having the same argument in different places in different functions or methods. Making this change was enabled by the previous item.

A few details

• Change argument order for Base.isapprox (Change argument order for Base.isapprox #13)

  Previously, I had Base.isapprox(::AbstractApprox, x, y). Now we use Base.isapprox(x, y, ::AbstractApprox). This order agrees with all other testing functions included in this package, such as, isone, isreal, ishermitian, etc.

• All predicates either extend Base/LinearAlgebra or are new to this package

  Previously, some predicates were copies of functions in Base/LinearAlgebra, but did not extend these functions with new methods, but rather were new functions under IsApprox. Other predicates in IsApprox had no function in Base/LinearAlgebra of the same name.

  This commit changes this so that all predicate functions of the same name as those in Base/LinearAlgebra now extend those functions.

  All predicate functions, including those extended from LinearAlgebra are exported.

  With this commit, you can do using IsApprox and suffer no name collisions with Base/LinearAlgebra.

  JET and Aqua find no ambiguities or piracy.

• This commit and the previous, precursor commit, include a breaking change to the API.

  Methods like Base.isapprox(::AbstractApprox, x, y) are now Base.isapprox(x, y, ::AbstractApprox) Methods like Base.isapprox(x, y; approx::AbstractApprox) are now Base.isapprox(x, y, ::AbstractApprox)

  In other words, the calling convention is now uniform.

Other changes

• Add tests and fix small bug

• Use 42.0 rather than 42 in isinteger test

  A method for rationalize(::Int) does not exist in older versions (at leat 1.6) of Julia. So we use a float instead so that the test runs on more versions.

• Fix bug in isunitary for Equal and IsApprox

  Don't compare result to zero. Add one and compare to one. This is the proper scale because we expect the vectors to be normalized.

• Explain new API

• Fix some incorrect statements in comments

  Base.isapprox does not use matrix norms. It treats multidimensional arrays as vectors.

  Closes Calling convention #12

• Try updating codecov action

  Added repository secret to GH repo config as well.

• Disable test using feature not in v1.0

• Add tests, remove two nonsense methods

  Looks like I tried to add methods to satisfy JET or Aqua in the past. But they don't have arg combinations that could ever be called, even because of ambiguity. I removed them and all tests, including JET and Aqua still pass.

[codecov]

Codecov Report

Attention: Patch coverage is 96.29630% with 3 lines in your changes missing coverage. Please review.

Project coverage is 94.77%. Comparing base (8cd8e66) to head (9418f50).
Report is 13 commits behind head on main.

Files with missing lines	Patch %	Lines
src/base_applications.jl	91.89%	3 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main      #14       +/-   ##
===========================================
+ Coverage   54.50%   94.77%   +40.27%     
===========================================
  Files           3        5        +2     
  Lines         200      249       +49     
===========================================
+ Hits          109      236      +127     
+ Misses         91       13       -78     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.