Add showdown status and card-reveal action to parsed hands

[ggratte]

Summary

Surfaces two new properties on every parsed hand, derived once in a post-parse pass so no site parser needs to change:

• HandHistory.WentToShowdown (bool) — true iff the hand reached showdown with at least two players still live.
• Player.RevealAction (enum: NotShown / MuckedAtShowdown / ShownAtShowdown / ShownVoluntarily) — what the rest of the table saw this player do with their hole cards.

Why

Until now, Player.HoleCards was a single nullable slot that parsers overwrote regardless of how the cards became known — dealt to hero, shown voluntarily, forced at showdown, or leaked via mucked [cards] in summary. Consumers couldn't distinguish a hero's dealt cards from a villain's showdown reveal, and had no convenience flag for WTSD.

The three facts are kept on orthogonal axes by design:

Question	Answered by
Does the parser have the cards?	Player.HoleCards != null
Is this player the hero?	HandHistory.Hero
What did the table see?	Player.RevealAction (new)

So a hero who reaches showdown and turns over cleanly gets RevealAction.ShownAtShowdown (table-visibility) without losing the fact that HoleCards was set from the "Dealt to" line (parser-knowledge).

Reveal values may be partial in Omaha / hi-lo split games (a player shows only some of 4/5/6 hole cards). This is documented on the enum; callers detect it by comparing HoleCards.Count against the expected count for GameDescription.GameType.

Implementation

• HandHistories.Objects/Cards/RevealAction.cs — new enum with XML docs.
• HandHistories.Objects/Players/Player.cs — new RevealAction property (defaults to NotShown).
• HandHistories.Objects/Hand/HandHistory.cs — new WentToShowdown property (defaults to false).
• HandHistories.Parser/Analyzers/ShowdownAnalyzer.cs — single linear walk of HandActions to populate both.
• HandHistories.Parser/Parsers/FastParser/Base/HandHistoryParserFastImpl.cs — invokes the analyzer right after FinalizeHandHistory in ParseFullHandHistory, so every site parser inherits the behaviour.

Test-infra fixes bundled in (required to run the suite on macOS)

These are small but worth flagging so reviewers know why they're in this PR:

• Retargeted both unit-test projects from netcoreapp3.1 to net8.0. .NET Core 3.1 is EOL (Dec 2022) with no native arm64 build, so the previous target couldn't run on Apple Silicon without x64 + Rosetta 2. Production libraries stay on netstandard2.0 / netstandard2.1 — downstream consumers unaffected. Rationale is inlined in both csproj files.
• Fixed a hard-coded \ path separator in SampleHandHistoryRepositoryFileBasedImpl.GetSampleHandHistoryFolder with Path.Combine. The old code returned null on unix filesystems, cascading into ~700 NullReferenceExceptions in tests that split the result.

After these fixes the suite goes from unrunnable on macOS → 994 passing / 27 failing. The remaining 27 failures are pre-existing cross-platform issues (Windows-1252 sample-file encoding read as UTF-8, and \r\n-only line splits on LF checkouts) and are unrelated to this change.

Test plan

• New ShowdownAnalyzerTests (5 tests) covers every enum value, the WentToShowdown flag, and the mucked-but-leaked-in-summary case — all pass.
• Full suite runs on macOS (net8.0): 994 passing / 27 pre-existing failures unrelated to this PR.
• CI / Windows run to confirm no regressions on the maintainer's primary platform.

🤖 Generated with Claude Code