docs: remove a spam token from a snapshot in one step

[kelsos]

Documents rotki/rotki#13257 on the snapshots page, for the 1.44.1 patch line.

New: Remove a high-value spam token

Replaces the old example, which walked through reconciling first and then deleting the row.

• Leads with marking the token as spam, which is usually enough: the list and the graph already leave ignored assets out.
• Warns against lowering locations or the net worth by hand around the token. The graph subtracts ignored assets from the stored total, so that makes the snapshot plot far below zero, which is how the reported user got stuck.
• The Totals do not match warning now offers Remove followed by the token's name, and says what is left to reconcile when the lowered values were typed a little off.
• Show them next to the excluded-value line, then delete the row.
• Split across locations with Fill for a token held in several locations, and why to fill the spam-only locations first.

It is now its own section, since it no longer goes through reconciling.

Brought in line with 1.44.1

These describe behaviour that changed in rotki/rotki#13243 and #13257:

• A stale stored total opens as an unsaved change that Save stores, instead of waiting for the next edit.
• Show spam alone reveals spam rows, and the hidden-rows chip says why rows are hidden and reveals them.
• Absorb difference into has no preselection any more, and greys out locations the difference would push below zero.
• The warning mentions the removal offer and links to the new section.

Checks

pnpm run lint, check:anchors, check:images and check:app-links pass. No images changed; the existing delete dialog screenshot still matches.

bugfixes was fast-forwarded to main before branching, so it already has the reconcile section from #147.