Hdict

Add Self: `_` for usage as parameter, argument, or value pointing to the hdict itself
Change semantics of _ for consistency:
_(1,2,3) = sample(1,2,3)
_["a"] = hdict["a"]

Fix apply for functions like `f = lambda a,b, *arg, c=1,d=2,e=13, **kwargs: 0`
Update hosh dependence
Now all fields are considered when calculating hdict hosh (including _* and *_)
Fix corner case for positional arg on apply
Add support for applying getattr() and calculate hosh of builtin functions and classes from str(function).encode()
Add support for None key, None meaning that the result of apply() is to be discarded.
Add hoshes property
Add option to hide types of DataFrame columns
Improve support for dataset operations: nom2bin, df2Xy

Finish test coverage
Finish pandas lazy loading by column
Improve output
Improve DF capabilities
Fix 'is Unevaluated' by "=="
Keep mirrors in ids list
Redesign algebra
Finish cache
Implementing cache, added mirror fields

Mostly rewritten for simplicity and reliability

Refactor and adopt sane syntax with mandatory 'apply'
Fix eq
Remove _id / _ids from 'd.data'
Turn mirror field optional; enabled explicitly by putting an "_" suffix for the original data

Fix show() for missing fields
Always try to apply()/solve pipeline
Fix _()
Improve missing field detector for pipeline
Disable callable field to avoid confusion
Impl _() as both hdict constructor and f application
Impl pipeline.show()
Allow missing field to trigger incomplete pipeline
Change syntax for consistence
_() now means hdict()
_._() now means apply()
_.f() now means apply()("f")

Handle circular refs
Rename to start_clone() finish_clone() for better semantics, and adopt abstractclass
Fix show() for set()
Make frozenhdict hashable