FSTree handles big headers incorrectly

[roman-khimov]

Expected Behavior

Store an object with header of MaxHeaderLen-1 size. GET/HEAD the object correctly.

Current Behavior

There will be a problem with some requests because FSTree APIs returns 16K from the beginning of the object and it's not the same as header (includes OID, signature and some protobuf overhead).

Possible Solution

Limit signature size (invocation script is 1K, verification script is also 1K in N3, that's a good start), fetch like 20K from storage.

Context

#3810 (comment)

Regression

Kinda. Before binary/streaming APIs it worked by always pulling whole object from the storage.

Your Environment

• Version used: 0.51.1

[cthulhu-rider]

func TestHeadOverflow(t *testing.T) {
	val := testutil.RandByteSlice(object.MaxHeaderLen - 49)

	var obj object.Object
	obj.SetID(oidtest.ID())
	obj.SetContainerID(cidtest.ID())
	obj.SetAttributes(object.NewAttribute("attr", string(val)))

	fst := setupFSTree(t, fstree.WithCombinedSizeThreshold(1<<10)) // for non-combined write

	require.NoError(t, fst.Put(obj.Address(), obj.Marshal()))

	_, err := fst.Head(obj.Address())
	require.NoError(t, err)
}

Error:      	Received unexpected error:
        	            	extract object stream from "/tmp/TestHeadOverflow4016725194/001/H/f/F/N/1KBiCKujKwaNUN6gUr3ydRGfrepGfS8c9ZW2iqkF.7LTnoRdt4EEbvJXWHZ3U2EQXUnTGPuxwuDDVZP1HtPDv": extract header and payload: invalid bytes field at offset 37: unexpected EOF