Better mechanism to control SetterFlags

首先跟踪 issue 测试示例中使用的 Python properties、_set_PMult setter、SetterFlags 和 end_edit()，然后比较所提议的 context-manager 和 default-property 方法将如何通过 C API 传递 flags。完成后应建立一个有文档说明的机制，支持单个操作和批量操作，包括 ImplicitSizes 和 AvoidFullRecalc，并使所示的基于 property 的用法能够正常工作。

Currently this is an advanced feature, and users need to leave the comfort of the Python properties, using instead the setter methods in order to pass both a value and the flags.

Some context, from the tests:

    alt.Clear()
    ls = alt.LoadShape.new('test_shape')
    with pt.raises(DSSException):
        ls.PMult = [1.0, 2.0, 3.0]

    ls._set_PMult([1.0, 2.0, 3.0], SetterFlags.ImplicitSizes)
    assert tuple(ls.PMult) == (1.0, 2.0, 3.0)

    ls.end_edit()

Some possible alternatives:

• Use a context manager in Python to keep the flags used in a block, use these flags on each Python function (when passing to the C API) and reset them later. This could be combined with the Edit context manager later.

with alt.with_setter_flags(SetterFlags.ImplicitSizes):
    ls = alt.LoadShape.new('test_shape')
    ls.PMult = [1.0, 2.0, 3.0]
    assert tuple(ls.PMult) == (1.0, 2.0, 3.0)
    ls.end_edit()

• Introduce a pair of API functions to control some default flags, expose it as a property. For this, we wouldn't need to change much of the Python code, but would require consuming the default flags on the engine:

    alt.DefaultSetterFlags = SetterFlags.ImplicitSizes
#...
    ls = alt.LoadShape.new('test_shape')
    ls.PMult = [1.0, 2.0, 3.0]
    assert tuple(ls.PMult) == (1.0, 2.0, 3.0)
    ls.end_edit()

Other examples include using SetterFlags.AvoidFullRecalc to minimize Yprim and SystemY changes for loads and generators. This is valid for both individual and batch operations.