Remove redundant encoding of syllabic segments/syllables from stress and tone enrichment

[mmcauliffe]

As noted by @massimolipari in #269

[massimolipari]

Unfortunately, this doesn't quite work. The stress/tone labels are correctly removed from the phone labels, but the syllable labels are left unchanged. (I imagine this is why syllables were re-encoded in the first place).

Here's some code to reproduce the issue (using the English test corpus):
stress_to_syll_test.py

The queries there generate this output:

node_type_phone_label: AH
node_type_phone_label: AE
node_type_phone_label: AO
node_type_phone_label: D
node_type_phone_label: S
node_type_phone_label: AA
node_type_phone_label: AO
node_type_phone_label: ER
node_type_phone_label: P
...
node_type_syllable_label: TH.R.AH1
node_type_syllable_label: S.T.IH0.NG
node_type_syllable_label: OW1
node_type_syllable_label: P.AH0.N
node_type_syllable_label: DH.AH1
node_type_syllable_label: P.R.AA1
node_type_syllable_label: P.ER0
node_type_syllable_label: EH1.N
node_type_syllable_label: T.R.AH0.N.S
node_type_syllable_label: AH0.V

Surely there must be some way to re-generate syllable labels without having to re-run syllabification?