stdlib: give Option and Result their combining half

[simontreanor]

Dogfooding finding #5, PR 5 of 5 (see #34) — the last sweep. Branched from main, independent of #46 (Map/Set) and #47 (String); all three can merge in any order.

Both types had map/bind/withDefault and the predicates, which handle one value at a time. Neither could combine two, or cross into another type without a match. Ten members: map2 orElse flatten iter toList exists for Option; map2 orElse iter toList for Result.

map2 is the one that earns its place. Two independent lookups that both have to succeed otherwise means nesting two matches:

Option.map2 (fun row col -> Square { row = row, col = col })
            (Map.tryFind r rows)
            (Map.tryFind c cols)

Option.map2 is Some only when both are. Result.map2 is Ok only when both are, and otherwise reports the first Error, so the earliest failure surfaces. There's a lowering test asserting a is checked before b, rather than trusting the reading.

orElse takes the fallback first, matching withDefault, so it reads as primary |> Option.orElse fallback under the pipe. Worth review: getting the order backwards would be silently wrong rather than a type error, since both arguments have the same type. That's why the argument order is stated on the member's own documentation line, not just here.

One test premise of mine was wrong, and the correction is instructive. I expected Result.map2 f (Ok 1) (Error 2) to be a type error, but it checks: Ok 1 leaves its error type free, so it unifies with anything. Two concrete error types are what conflict, and the test pins that instead.

Tests: 7 typecheck cases (combining, the orElse order, one-level flatten, the shared error type, effect-carrying iter), 1 lowering assertion (first-error ordering), and 2 end-to-end programs covering all ten members. Full suite, clippy and fmt clean.

With this the sweep is complete: 94 members across List, Seq, Set, Map, String, Option, Result, plus fst/snd, all documented with their complexity. Next is the FSharp.Core audit pass, then finding #7's import diagnostic.