Refactor wallet code to eliminate most `plan::foo` methods in favor of the Planner API

[hdevalence]

Is your feature request related to a problem? Please describe.

Now that we have the Planner API, we have a bunch of methods like

/// Generate a new transaction plan delegating stake
#[instrument(skip(fvk, view, rng, rate_data, unbonded_amount, fee, source_address))]
pub async fn delegate<V, R>(
    fvk: &FullViewingKey,
    view: &mut V,
    rng: R,
    rate_data: RateData,
    unbonded_amount: u64,
    fee: Fee,
    source_address: Option<u64>,
) -> Result<TransactionPlan>
where
    V: ViewClient,
    R: RngCore + CryptoRng,
{
    Planner::new(rng)
        .fee(fee)
        .delegate(unbonded_amount, rate_data)
        .plan(view, fvk, source_address.map(Into::into))
        .await
        .context("can't build delegate plan")
}

which is just a worse wrapper around the Planner API.

Describe the solution you'd like

All of these methods, with the exception of sweep, should be eliminated, making sure that the planner API is at least as ergonomic as the method it's eliminating. Alternatively, if there's some reason we can't eliminate one, we should have a reason why it doesn't fit the planner API. (For instance, we can't eliminate sweep, because it makes a list of transaction plans, while the planner API does a single transaction, so that's a differently shaped problem).

[hdevalence]

When we do this, we should eliminate the wallet crate entirely. It's a vestigial holdover that currently serves no purpose.