wallet: wire -sendto into main(), where it never was

[Bitflash-sh]

What is missing

CmdSendTo() has lived in walletcmd.cpp since #115, and nothing ever declared or called it. The flag is dead code: bitflash -sendto=ADDRESS,AMOUNT starts a node and says nothing, because an unrecognised argument is not an error in this tree.

The v1.2.14 notes announce it under "A headless node can finally spend" — and the headless Linux binary is precisely the one that does not have it. -help on the published bitflash-node lists /newaddress and stops.

The published Windows binary does have it. It was built from a working tree carrying these two files uncommitted, so the two artifacts of the same release disagree with each other and neither matches the tag.

What this changes

Only the wiring that was missing:

• the declaration in walletcmd.h
• the dispatch in main()
• the -help line

No change to CmdSendTo itself.

Testing

Built from this branch; all five self-tests pass. Then exercised the flag against a throwaway data directory, since the whole failure was that it silently did nothing:

command	result
-help	lists /sendto=ADDRESS,AMOUNT
-sendto=NOTANADDRESS,1	Not a valid address: NOTANADDRESS, exit 1
-sendto=<valid>,-5	Not a valid amount: -5, exit 1
-sendto=<valid>,999	refused on an empty wallet, exit 1

Every refusal happens before the wallet is touched, which is what CmdSendTo already promised and could never demonstrate.

Not in this PR

BTF_ACCEPT_THREADS was the third file uncommitted in that tree, set to 3. It stays at 1 here. The published Windows binary is running with 3, which the v1.2.14 notes say was deliberately left out; that is a separate problem and is not fixed by this change.

[Bitflash-sh]

Thank you for catching this, and the title is exactly right — it never was wired.

For the record of how it happened, because the shape of the mistake matters more than the missing lines: I wrote CmdSendTo on a branch, tested it against a live wallet, and then moved to the next fix without ever committing it. Switching branches carried the uncommitted files along, and a later git add src/walletcmd.cpp for an unrelated change swept the function body in with it. The header and the main() dispatch stayed behind.

So two things went wrong at once. #115 shipped a diff wider than its description — the same contract violation I had corrected on #112/#113 a few hours earlier — and v1.2.14's release notes promised a command that build cannot run. I have added a correction at the top of that section in the v1.2.14 notes pointing at 1.2.15, since anyone reading them on the old release would otherwise be told something false.

The lesson I am keeping: a feature is not done when it works in my shell, it is done when it is committed on its own branch and the diff matches what the description says.