feat: add --send to contract invoke and default to only sending transaction if there are writes in the ledger footprint

[willemneal]

What

See initial discussion here: #225

Why

Currently users must opt in to not sending a transaction with --is-view. This deprecates --is-view and flips the default to only send if there are writes in the ledger footprint returned by the simulation.

It improves UX because now when making a view call a user must wait for the transaction be complete vs viewing the simulation result.

Close #225

Known limitations

[TODO or N/A]

[leighmcculloch]

Couple requests inline.

I think the concept is still pending feedback from @janewang and others.

But I've been hoping to see this functionality for a couple years (#225) so I think adding this or something like this is 💯.

[willemneal]

@leighmcculloch should rename to use your issue's names? say slack ❤️

Thanks for the review!

[leighmcculloch]

Took another look, one fix needed, and otherwise just needs events 👏🏻

[leighmcculloch]

Looks good. One thing missing, an env var for the option, otherwise lgtm.

[leighmcculloch]

I think we need a follow up change to this change prior to this change being released. (cc @janewang @fnando).

In this change we didn't take into the account the implication of txs that are read-only (no storage writes, no events), but that require auth. Auth checks are never run, and auth code is not executed.

It's not obvious what the behavior should be.

Simulating and considering the tx as read-only may be appropriate. Or it maybe surprising.

Maybe we at least need a warning outputted that the tx is not being sent because simulation indicated that the tx was read-only.

Or maybe we need to treat any tx with an auth as a read-write. Folks can still force it not to submit with --send=no.

Thoughts anyone?