[tts-adapter-sapinative] re-unify microsoft voices native adapters

[NPavie]

Given the issues encountered with breaking changes between Windows updates with the WinRT runtime, this PR aims to reunify the Onecore and SAPI native adapters in favor of using the legacy SAPI library to also connect to onecore voices. This should moreover simplify maintainability given the complexity induced by WinRT/C++ coding and build process.

Changes summary

• Dropped the Onecore WinRT native adapter and tests, which can break between Windows updates if not rebuilt with updated Windows SDK
• Add Onecore voices usage through legacy SAPI
• Improve SAPI marks information retrieval to use
• Extended the timeout of the adapter for experimenting with natural voices longer synthesis time

Notes

Third party voices and tools like NaturalVoicesSAPIAdapter also rely on legacy SAPI to expose natural or online voices, as Microsoft does not officially provide documentation on how to install third party voices in Onecore voices set.

Also note that support of voices exposed by NaturalVoicesSAPIAdapter cannot be fully considered, given the following problems encountered while testing it with the fixes included by this PR :

• Some issues can still be encountered with marks reporting of natural voices with no clear cause identified yet
• Online (Edge) voices don't seem to report any marks with the websocket adapter included in NaturalVoicesSAPIAdapter

[bertfrees]

@NPavie Nice work. Thank you! The is one issue with duplicate code. If that is fixed we can merge.

[bertfrees]

@NPavie Thanks!