Fixedmap optimizations

[Scooletz]

This PR optimizes FixedMap component internal encoding of data, so that more entries can be packed into a single map. This PR lowers the disk space by total 10%.

Changes introduced:

1. The length of size reduced to one byte. The size of length reduced from 2 bytes to 1 byte as no entry will ever be more than 255 bytes long. This gives savings ~5%.
2. The prefix extraction and encoding. This removes 1 or 2 bytes per entry as well, but required crazy unsafe ref magic over enumerator.

Before

Using in-memory DB for greater speed.
Initializing db of size 18GB
Starting benchmark with commit level FlushDataOnly
Preparing random accounts addresses...
Accounts prepared

(P) - 90th percentile of the value

At Block        | Avg. speed      | Space used      | New pages(P)    | Pages reused(P) | Total pages(P)
           5000 |  892.0 blocks/s |          2.35GB |            1606 |             188 |            1733
          10000 |  713.3 blocks/s |          3.58GB |            1917 |             171 |            1962
          15000 |  684.2 blocks/s |          4.06GB |            2042 |             157 |            2063
          20000 |  661.6 blocks/s |          4.31GB |            2111 |             143 |            2121
          25000 |  639.4 blocks/s |          4.60GB |            2147 |             129 |            2163
          30000 |  620.2 blocks/s |          5.42GB |            2181 |             116 |            2219
          35000 |  564.8 blocks/s |          7.38GB |            2205 |             127 |            2293
          40000 |  523.8 blocks/s |         10.37GB |            2241 |             158 |            2385
          45000 |  499.8 blocks/s |         13.79GB |            2299 |             177 |            2475
          49999 |  489.7 blocks/s |         17.13GB |            2365 |             184 |            2545

Writing state of 1000 accounts per block, each with 1 storage, through 50000 blocks, generated 50000000 accounts, used 17.13GB

Reading and asserting values...
Reading state of all of 50000000 accounts from the last block took 00:00:35.0334893
90th percentiles:
   - new pages allocated per block: 0
   - pages reused allocated per block: 1239
   - total pages written per block: 1244

After one-byte length

Using in-memory DB for greater speed.
Initializing db of size 18GB
Starting benchmark with commit level FlushDataOnly
Preparing random accounts addresses...
Accounts prepared

(P) - 90th percentile of the value

At Block        | Avg. speed      | Space used      | New pages(P)    | Pages reused(P) | Total pages(P)
           5000 |  873.9 blocks/s |          2.34GB |            1605 |             187 |            1730
          10000 |  705.7 blocks/s |          3.58GB |            1915 |             170 |            1960
          15000 |  679.5 blocks/s |          4.05GB |            2041 |             156 |            2063
          20000 |  656.4 blocks/s |          4.29GB |            2107 |             142 |            2119
          25000 |  629.4 blocks/s |          4.54GB |            2143 |             129 |            2155
          30000 |  613.4 blocks/s |          5.19GB |            2177 |             116 |            2207
          35000 |  541.7 blocks/s |          6.88GB |            2199 |             118 |            2273
          40000 |  487.9 blocks/s |          9.70GB |            2231 |             151 |            2365
          45000 |  486.9 blocks/s |         13.11GB |            2287 |             174 |            2457
          49999 |  496.1 blocks/s |         16.56GB |            2353 |             184 |            2535

Writing state of 1000 accounts per block, each with 1 storage, through 50000 blocks, generated 50000000 accounts, used 16.56GB

Reading and asserting values...
Reading state of all of 50000000 accounts from the last block took 00:00:34.3674436
90th percentiles:
   - new pages allocated per block: 0
   - pages reused allocated per block: 1239
   - total pages written per block: 1244

After prefix extraction

Using in-memory DB for greater speed.
Initializing db of size 18GB
Starting benchmark with commit level FlushDataOnly
Preparing random accounts addresses...
Accounts prepared

(P) - 90th percentile of the value

At Block        | Avg. speed      | Space used      | New pages(P)    | Pages reused(P) | Total pages(P)
           5000 |  771.9 blocks/s |          2.30GB |            1593 |             185 |            1720
          10000 |  672.3 blocks/s |          3.55GB |            1909 |             169 |            1954
          15000 |  634.6 blocks/s |          4.03GB |            2036 |             154 |            2057
          20000 |  581.8 blocks/s |          4.28GB |            2103 |             141 |            2115
          25000 |  610.8 blocks/s |          4.51GB |            2143 |             128 |            2155
          30000 |  595.9 blocks/s |          5.01GB |            2175 |             115 |            2199
          35000 |  535.0 blocks/s |          6.42GB |            2195 |             110 |            2257
          40000 |  503.5 blocks/s |          8.98GB |            2223 |             140 |            2341
          45000 |  471.7 blocks/s |         12.18GB |            2273 |             165 |            2433
          49999 |  464.2 blocks/s |         15.48GB |            2337 |             175 |            2509

Writing state of 1000 accounts per block, each with 1 storage, through 50000 blocks, generated 50000000 accounts, used 15.48GB

Reading and asserting values...
Reading state of all of 50000000 accounts from the last block took 00:00:33.6083939
90th percentiles:
   - new pages allocated per block: 0
   - pages reused allocated per block: 1239
   - total pages written per block: 1244

[github-actions]

Package	Line Rate	Branch Rate	Health
Paprika	89%	77%	✔
Summary	89% (1151 / 1299)	77% (275 / 357)	✔

Minimum allowed line rate is 75%