Convert UniqueID::nil() to a constructor

[stephanie-wang]

What do these changes do?

The default UniqueID() does not initialize the ID data, which can make statements like ObjectID object_id; error-prone. This merges the UniqueID::nil() static method into the default constructor.

[pcmoritz]

LGTM conditionally on tests (esp. linting) passing.

[robertnishihara]

This looks good to me.

[pcmoritz]

above we used .is_nil() for this, we might want to unify for consistency

[stephanie-wang]

Thanks!

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/10106/
Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/10107/
Test FAILed.

[raulchen]

Actually I prefer keeping the UniqueID::nil() method, for 2 reasons:

1. compared to UniqueID(), the meaning of UniqueID::nil() more clear. Users know this method returns a nil ID, without needing to refer to the document.
2. With factory method UniqueID::nil(), we can actually cache one NIL instance for all use cases (assuming UniqueID is immutable). Compared to creating a new instance every time, this is more efficient.

But, it's good to initialize data to nil in the constructor.

[stephanie-wang]

Thanks, @raulchen. I left in UniqueID::nil() method but changed it to return a reference to a static variable.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/Ray-PRB/10128/
Test PASSed.