Always try to resolve conversion for Any type

[JaroslavTulach]

Pull Request Description

Fixes #5898 by removing Catch.panic and speeding the sieve.enso benchmark from 1058 ms to 514 ms. Should there be no dedicated conversion, let's use one defined on Any type - e.g. defining a conversion from(Any) makes such a conversion is always available.

Checklist

Please ensure that the following checklist has been satisfied before submitting the PR:

• All code follows the
  Java,
• All code has been tested:
  • Unit tests have been written where possible.

[radeusgd]

Looks good.

Maybe worth adding a few test cases for it?

I'd imagine at least 3:

1. Have a conversion both on type Foo and Any: verify that for an instance of type Foo the more specific one (Foo, not Any) is selected.
2. Have a conversion on Any but not Foo - check that Any is selected - essentially what this PR does and is implicitly tested by the == not breaking, but I think it's good to have an explicit test for it.
3. Have no conversion defined on neither Foo nor Any - to ensure that the conversion resolution error is still correctly reported.

[JaroslavTulach]

Maybe worth adding a few test cases for it?

There were fifteen failing tests in test/Tests project when I changed the implementation of Any.== - I had to fix them all. E.g. while no test was added, the code is certainly tested.

[radeusgd]

Maybe worth adding a few test cases for it?

There were fifteen failing tests in test/Tests project when I changed the implementation of Any.== - I had to fix them all. E.g. while no test was added, the code is certainly tested.

Sure, but these explicit tests would also serve as documentation on what is the behaviour we currently expect.

[JaroslavTulach]

Maybe worth adding a few test cases for it?

There is a lot of tests in Conversion_Spec.enso including those that you propose:

Have a conversion both on type Foo and Any: verify that for an instance of type Foo the more specific one (Foo, not Any) is selected.

Not_Foo type & its tests verify this.

Have a conversion on Any but not Foo - check that Any is selected - essentially what this PR does and is implicitly tested by the == not breaking, but I think it's good to have an explicit test for it.

That's "should be able to convert Any" test, I'd say.

Have no conversion defined on neither Foo nor Any - to ensure that the conversion resolution error is still correctly reported.

There are two "should fail graciously..." tests that seem to cover this behavior.

At the end I am just adding two tests for foreign Java object and a type: 40cc46b

[JaroslavTulach]

Running benchmarks for this PR. Manual comparison with develop branch shows no slowdowns, and indicates few speedups: