test: analyse_later_board TT counter reset/report (#379 item 4)

[wopdevries]

Adds the reset/report lifecycle for analyse_later_board(), item 4 of #379 and the last remaining item.

Change (library/src/solver_if.cpp)

analyse_later_board() reuses the caller's warm SolverContext/TT across a whole AnalysePlayBin call. It previously had no counter reset and no DDS_PRINT_TT_STATS report at all.
It now resets tt_lookup_count, tt_hit_count and the TT op-stats at entry. Only the counters are reset. The warm TT is deliberately left untouched, per the #157 warm-context contract.
It reports on both the normal exit and the cardCount<=4 last-trick early return, which previously reported nothing.
The reporting now lives in one shared print_tt_stats() helper, including the n/a TransTableS case from #393. solve_board_internal, solve_same_board and analyse_later_board all call it, replacing the two duplicated inline blocks.

Tests (library/tests/solve_board/analyse_later_board_tt_stats_test.cpp)

ResetsLookupAndHitCountersIndependentlyOfPriorSolve seeds tt_lookup_count and tt_hit_count with a large sentinel just before a direct analyse_later_board call. It confirms the printed counts can't include the sentinel.
ResetsTtOpStatsIndependentlyOfPriorSolve covers op-stats, which have no public setter to poison. It compares two identical runs, one with an extra manual reset_op_stats() before the call. If analyse_later_board really resets, both print identical adds, overwrites and harvests. This doesn't depend on how much work the later search does. It forces DDS_TT_KIND=large so the counts are real numbers, not n/a.
ReportsOnBothNormalAndLastTrickEarlyReturnExits drives a full play through the real AnalysePlayPBN on a forced one-suit-per-hand deal. It confirms every card, including the last-trick early return, gets its own report.

Verification

Each of the three resets (lookups, hits, op-stats) was disabled individually. Each break is caught only by the test meant to catch it.
The tests also fail against the pre-fix code.
#402's tt_print_stats_test still passes against the refactored solve_board_internal and solve_same_board.
Both test targets pass under bazelisk.

Closes #379.

[Copilot]

Copilot review overview

🟡 Changes recommended

Address the requested helper reuse, complete counter assertions, and naming consistency.

Review effort: Lite
Findings: 3

Open (3)

• Replace duplicated TT stats blocks with the shared helper · New
• Add assertions covering TT hit and operation counter resets · New
• Rename test helpers to snake_case · New

What changed in this PR

Adds TT counter reset/reporting to analyse_later_board() with coverage for normal and last-trick exits.

Changes:

• Adds TT statistics reporting and counter resets.
• Adds deterministic lifecycle tests.
• Registers the new Bazel test target.

File	Summary
library/​src/​solver_if.cpp	Implements TT counter lifecycle and reporting.
library/​tests/​solve_board/​analyse_later_board_tt_stats_test.cpp	Adds reset and reporting integration tests.
library/​tests/​solve_board/​BUILD.bazel	Registers the test target.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🔵 Needs a closer look

Make the op-stat reset assertion deterministic.

Review effort: Lite
Findings: 1

Open (1)

• Add assertions covering TT hit and operation counter resets

Resolved since last review (2)

• Rename test helpers to snake_case
• Replace duplicated TT stats blocks with the shared helper

[Copilot]

Copilot review overview

🟢 Approval recommended

All reviewed changes are covered and no unresolved blocking issues were identified.

Review effort: Lite
Findings: 1

Open (1)

• Add assertions covering TT hit and operation counter resets

[wopdevries]

Addressed in 9ae71e9 and f1eff63. Hits are covered by the sentinel check, and op-stats are covered by comparing two runs that differ only in a manual reset_op_stats(), so it doesn't depend on how much work the later search does.
print_tt_stats() is now shared by solve_board_internal, solve_same_board and analyse_later_board, since Copilot asked for that dedup and it changes code beyond analyse_later_board.