slicer: Do not write split ID into parent header

[cthulhu-rider]

Split ID relates to split-chain only, original root object should not carry this information, otherwise, it will be perceived by the system itself as part of a sliced chain that does not exist.

[smallhive]

I consider you should discuss this with @carpawell, because we fixed/added the splitID to first header here #448, to fix some bug inside node

[carpawell]

@smallhive, the first child obj and the root obj are not the same so that PR is correct.

@cthulhu-rider, do we have a test for "every child has a SplitID"? Slicer code is quite complex such a pair of tests would help (if not yet): (1) every part has a correct and full Split header and (2) a root object has no split info ((2) is already done in that PR).

BTW, is our API description full enough to understand that we should not add splitID to the root object?

[cthulhu-rider]

@carpawell

do we have a test for "every child has a SplitID"?

yep, i modified place which checks case 1.

neofs-sdk-go/object/slicer/slicer_test.go

Lines 561 to 564 in bc6b51e

	for _, h := range writer.headers {
	splitID := h.SplitID()
	require.NotNil(t, splitID)
	require.Equal(t, writer.splitID.ToV2(), splitID.ToV2())

API description full enough to understand

full enough? guess no. Strictly speaking, root object is not a part of split-chain cuz it's an autonomous object. When we restore original object, we don't expect to see any split headers, huh? I think so.

if you ask me why this (root split ID absence) is not a recommendation, but a requirement, I will not give an exact answer. There is an open issue nspcc-dev/neofs-node#2451