wallet: make wallet.dat a file you can actually back up

[Bitflash-sh]

Fixes the first half of #40, where the reporter believes they lost funds after backing up wallet.dat on its own — the one file anybody would think to save.

The chain

Working from the bottom up, because the wallet symptom is the last link:

1. Shutdown aborted. _endthread() is pthread_exit() on POSIX, and glibc implements that by throwing abi::__forced_unwind through the stack. It is not a real exception; catching it without rethrowing is a fatal error by contract. CATCH_PRINT_EXCEPTION ended in a bare catch(...), so every Linux shutdown ended:

StopNode()
UNKNOWN EXCEPTION  bitflash-node in ThreadSocketHandler()
FATAL: exception not rethrown          ← abort, exit 134

2. So nothing after StopNode() ran. Which mattered because —

3. Berkeley DB welds files to environments. Every page carries a log sequence number belonging to the environment that wrote it. Carried elsewhere those numbers point at logs that do not exist, and BDB refuses to write: Db::put: Invalid argument. DBFlush(true) calls lsn_reset, which is exactly the cure — and DBFlush has been declared since 2009 and called from nowhere in this tree. Even had it been called, the abort stood in front of it.

4. So the wallet was never released from its directory, and the one file users copy was the one file that could not travel.

Three fixes, three different cases

	Covers
Rethrow abi::__forced_unwind, call DBFlush(true) on both shutdown paths	New wallets — left portable by construction
Adopt orphaned files on open: a freshly created environment beside pre-existing .dat files means they came from elsewhere, so reset their sequence numbers	Wallets already stranded by older builds, including the reporter's if they still hold the file
BackupWallet(), reachable as /backupwallet=FILE — checkpoint, copy, clear the copy's sequence numbers	Taking a backup at all. Bitcoin 0.1.0 had a backup command; this tree lost it in the wxWidgets removal and has shipped with no way to take one since

DBFlush's own txn_checkpoint and lsn_reset are wrapped too — they throw, and throwing on the way out is how this started.

Verification

Both platforms:

• Shutdown exits 0, was 134
• DBFlush(true) runs — previously 0 times, ever
• A lone wallet.dat in an empty directory opens with 0 exceptions, logging CDB() : adopted wallet.dat from another environment
• Key records byte-identical before and after the reset — defaultkey, key, name, setting, version, same digest
• /backupwallet output reopens standalone in a fresh directory; a bad destination reports an error and writes nothing, no abort

Before, on the same file:

Loading wallet... → Height=0 →
terminate called after throwing an instance of 'DbException'
  what(): Db::put: Invalid argument

Not addressed here

The larger half of #40. Keys are generated as needed rather than derived from a seed, so any backup is a point-in-time snapshot and coins paid to addresses created after it are not recoverable from it. /backupwallet says so on every run, and the README change says so too, but the real answer is deterministic derivation — a wallet format change and its own piece of work, worth its own issue rather than being smuggled in here.

One known limitation: the Windows GUI binary is linked -mwindows, so /backupwallet prints to debug.log rather than the console. Pre-existing, not introduced here.

[Bitflash-sh]

Pushed one commit after opening this: the README section on backups.

Flagging it rather than letting it land quietly, because the description above already said "the README change says so too" — and it did not. I wrote that before writing the section and never checked. The push makes the claim true instead of leaving it false; it is the only change since the PR opened, and it touches no code.