Extract a sync sibling of run_host_async for ores.cli/ores.shell

Add a sync counterpart to ores::service::service::run_host_async() (projects/ores.service/include/ores.service/service/host_runner.hpp) for ores.cli=/=ores.shell, whose host.cpp bodies are correctly not on the async helper (neither has an io_context — see the investigation this was filed from) but still duplicate ~70% of their own structure with each other: parse args -> check for empty optional<config> -> init lifecycle_manager -> log args/config -> try { construct application; run; } catch (const std::exception&) { diagnostic_information-or-e.what() dance; rethrow; }. Since neither tool's application is uniformly constructible (ores.cli needs (output_stream, db_opts), ores.shell is already all-default-optional), the helper needs to take an application-construction callable (cfg -> Application) rather than assuming default-constructibility like the async helper does, plus an optional early-exit hook for ores.cli's DB-free ore_roundtrip branch.

Closes the remaining boilerplate-duplication gap the async extraction (PR #1855) deliberately left out — these two tools were correctly excluded from run_host_async (sync-vs-async is genuinely load-bearing, not historical — see the investigation this followed from), but that doesn't mean there's no shared-code opportunity left; it's just a different, sync-shaped envelope than the one already extracted.

• projects/ores.cli/src/app/host.cpp
• projects/ores.shell/src/app/host.cpp
• projects/ores.service/include/ores.service/service/host_runner.hpp

• Review whether ores.cli/ores.shell's host.cpp divergence is load-bearing or historical — the investigation that surfaced this
• Unify application.cpp and service initialisation across services — the original async extraction