Report recovered coroutine panics with Go stacks

[skhaz]

Summary

• include debug.Stack() in recovered non-API coroutine panic errors when IncludeGoStackTrace is enabled
• report the same recovered panic text through an optional PanicHandler
• contain reporter panics so the original VM error remains valid

Validation

• go test -timeout 120s ./... with Go 1.23.12
• go test -race -timeout 120s ./... with Go 1.23.12
• golangci-lint v2.8.0 with the CI lint set
• all three CI fuzz targets for 60 seconds each
• independent code review with no remaining findings

Context

This complements the coroutine yield fix in #40. It makes recovered Go panics diagnosable by both callers and host logs.