types: fix getblockstats doc for v21 and later

[satsfy]

Docs are wrong in getblockstats txs field.

The txs field counts all transactions in the block, including the coinbase. Core's source passes block.vtx.size() directly, which includes the coinbase transaction at index 0. Bitcoin Core doc agrees.

Example, I just called

bitcoin-cli -regtest generatetoaddress 101 (bitcoin-cli -regtest getnewaddress)

Now I run

bitcoin-cli -regtest getblockstats 50 | grep "\"txs\""
  "txs": 1,

Returns 1 because that is the coinbase transaction.

Also, swtotal_weight and total_weight fields say "divided by segwit scale factor (4)" but it was dropped from the Core docs at v21 and should not appear in v21 or later.

[jamillambert]

I checked the docs and it looks like this changed in v20, i.e. v17-19 say (excluding coinbase). I didn't check if it actually changed or if the docs were fixed. I say this because I know of other cases where the docs in v17 or so were incorrect.

[satsfy]

Great catch, my bad for letting that slip

So in version v20 it says excluding.
In version v21 it says including.

Also fix swtotal_weight and total_weight in and later because the phrase "divided by segwit scale factor (4)" was dropped from the Core docs at v21 and should not appear in v21+.

[jamillambert]

So in version v20 it says excluding. In version v21 it says including.

I just checked again to see why we get a different version where this changed and it changed between 20.1 and 20.2. We use 20.2 in types.

[tcharding]

I read the PR description and it reads like this is a docs only fix but then the patch diff is big and there are loads of code changes. Is this PR introducing new types just to fix the docs? If so this is a policy change for the repo and cannot be done lightly. Looks like this is tackling the problems outline in #457 and #59 but only partially. If so we need a more principled approach.