Write Widen arrays through an array plugin - #11002

Tracking Issue: #10701

Bottom of the decimal `Widen` stack. #11002 builds on it. No release
gate.

## Summary

A plugin can need another encoding's wire form without taking over its
reader. `serialized_ids()` currently serves both jobs, so a `Widen`
writer cannot emit a narrow decimal node while preserving the decimal
reader.

## Changes

Reader registration now uses `read_ids()`, while serialization validates
its output against `write_ids()`. A writing session does not need a
reader for every ID it writes. Existing plugins keep their current IDs,
and tests cover reader ownership and the two validation contracts.
`LayoutPlugin` is unchanged.

## API Changes

Replaces `ArrayPlugin::serialized_ids()` with required `read_ids()` and
`write_ids()` methods, with no defaults. Read IDs stay ordered oldest to
newest.

---------

Signed-off-by: Connor Tsui <[email protected]>

Tracking Issue: #10701

Based on `develop`, including the writer work previously in #11002.
#11003 and #10994 build on it. No release gate.

## Summary

Canonical decimal width follows precision, so small values under
`Decimal(38, 2)` still occupy 16 bytes each. A decimal `Widen` keeps
those values in a lower-precision child while retaining the column's
logical dtype.

## Changes

The child precision must hold every valid value and have a narrower
canonical width. Execution restores the logical width. Until the later
kernels land, operations use that canonical fallback.

Serialization writes a concrete decimal leaf through the existing
`vortex.decimal` wire form without executing values or validity. Integer
`Widen` arrays, non-decimal leaves, and nonserializable validity require
explicit preparation by the caller. Its metadata stores the width, while
the containing format supplies the parent precision. This keeps an `i32`
child at four bytes per value, including IPC and writes without a
compressor. A serialized-byte regression test checks that size
independently of the eventual read encoding.

<details>
<summary>Historical local benchmark results</summary>

These retained measurements predate this style pass and rebase. They do
not measure the current head.

`cargo bench -p vortex-array --bench widen_decimal`, Apple M4, medians.
The numbers were measured at the top of this stack before the split, so
they cover the whole stack, not this layer alone. The baseline is the
canonical input, and the candidate is the same values as a `Widen`
(child precision 5 under precision 10, 9 under 18, and 18 under 38).
CodSpeed on this PR measures the layer itself.

| bench | p10 canonical | p10 Widen | p18 canonical | p18 Widen | p38
canonical | p38 Widen |
| --- | ---: | ---: | ---: | ---: | ---: | ---: |
| `filter` | 22.16 µs | 5.083 µs | 13.66 µs | 4.916 µs | 10.58 µs |
24.74 µs |
| `execute` | 10.59 ns | 2.832 µs | 10.26 ns | 2.926 µs | 10.34 ns |
9.895 µs |

`execute` on a `Widen` is the widening copy, which the canonical input
does not need. The canonical `filter` medians at p10 and p18 were
bimodal across runs, and `filter` at p38 is slower on the `Widen`.

</details>

## API Changes

`WidenArray::try_new` accepts decimal children. Adds
`WidenArray::encode_decimal` and registers a writer-only `WidenPlugin`,
whose only write ID is `vortex.decimal`.

## Validation

Source review and rebase range-diffs only. No local tests, builds,
linting, or documentation checks were run. A commit hook flagged one
formatting wrap, corrected manually without rerunning the check.

---------

Signed-off-by: Connor Tsui <[email protected]>