fix: read only nodes along the path on non-existent key deletion

[Longarithm]

Yet another cornercase fix: if deleted key doesn't exist, read only nodes required to prove its non-existence, thus only nodes along the path.
The chaos it caused is described on #11065 and discussed on Zulip.
This is not a protocol change, because currently nodes read on Trie::update don't impact costs, so we can modify current disk trie logic to follow the principle that only necessary nodes should be read.

Unfortunately disk tries logic, on descending, always subtracts child' memory usage from parent' memory usage and requires "backwards iteration" along the path to recompute potentially changed memory usages. I want to minimise changes, so I just introduce key_deleted to compute memory usages but avoid reads caused by node restructuring if key wasn't found. If I just return early, I get attempt to subtract with overflow on calc_memory_usage_and_store.

Memtrie logic is already correct because it computes memory usages after processing all keys, which is cleaner.

I also restructure trie consistency changes. Now these tests:

• always use memtries by default, as this is stateless validation flow we want to test;
• always include comparison of partial storages generated by disk tries, memtries and recorded storages;
• use random probabilities and any insert/delete operations for existing/missing keys. Unfortunately with previous impl all test still can pass, but in like 9/10 runs at least some of them fails. These tests are random, however, so I don't target zero flakiness.

Also tests with/without flat storage are better parametrised.

[codecov]

Codecov Report

Attention: Patch coverage is 93.77778% with 14 lines in your changes are missing coverage. Please review.

Project coverage is 71.06%. Comparing base (4d506a7) to head (d3bedf8).
Report is 11 commits behind head on master.

Files	Patch %	Lines
core/store/src/trie/insert_delete.rs	82.60%	8 Missing and 4 partials ⚠️
core/store/src/trie/trie_recording.rs	98.34%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #11071      +/-   ##
==========================================
- Coverage   71.14%   71.06%   -0.08%     
==========================================
  Files         761      763       +2     
  Lines      152857   153082     +225     
  Branches   152857   153082     +225     
==========================================
+ Hits       108748   108791      +43     
- Misses      39672    39851     +179     
- Partials     4437     4440       +3     

Flag	Coverage Δ
backward-compatibility	0.24% <0.00%> (-0.01%)	⬇️
db-migration	0.24% <0.00%> (-0.01%)	⬇️
genesis-check	1.43% <0.00%> (-0.01%)	⬇️
integration-tests	36.82% <19.11%> (-0.02%)	⬇️
linux	69.49% <93.77%> (-0.13%)	⬇️
linux-nightly	70.55% <93.77%> (-0.09%)	⬇️
macos	54.16% <93.75%> (+1.52%)	⬆️
pytests	1.66% <0.00%> (-0.01%)	⬇️
sanity-checks	1.44% <0.00%> (-0.01%)	⬇️
unittests	66.74% <93.75%> (-0.06%)	⬇️
upgradability	0.29% <0.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.