### Description of the issue/feature this PR addresses:
This PR introduces batch processing to the VAT Books ES (Libros de IVA) report generation. When attempting to export the report for periods containing a massive volume of invoices, the ORM cache continuously accumulates records, leading to severe memory consumption. By implementing batching and explicitly clearing the environment cache, use memory use will remain stable and efficient.

### Current behavior before PR:
Generating the VAT Books report loads all account move lines into memory at once. Because the ORM cache is never cleared during the iteration, RAM usage spikes continuously. On databases with tens or hundreds of thousands of invoices in a single period, this leads to significant performance degradation, worker timeouts, or complete Out-Of-Memory (OOM) crashes.

### Desired behavior after PR is merged:
The report engine now splits the recordset into manageable batches (e.g., 50,000 accounts per batch). After processing each chunk to extract the income and expense line values, invalidate_model() is called to flush the ORM cache related to the searched records. This frees up memory continuously, keeping the server's RAM usage flat and allowing the successful export of massive datasets without crashing.

### Benchmark:
The model is iterating through ~1.1M account move lines when generating the full report. 

For Memory:
| # Input Data | Before PR | After PR|
| -------- | -------- | -------- |
| ~77,000 account move lines | 385 MB | 666 MB |
| ~340,000 account move lines |1.2 GB | 1.5 GB |
| ~1.2M account move lines | MemoryError | 1.5 GB |

For Speed:
| # Input Data | Before PR | After PR|
| -------- | -------- | -------- |
| ~77,000 account move lines | 32s | 12s |
| ~340,000 account move lines | 2:29min | 1:11min |
| ~1.2M account move lines | MemoryError | 4:11min |

### Reference

opw-6037414

-----------------------------------------------------------------
I confirm I have signed the CLA and read the PR guidelines at [www.odoo.com/submit-pr](http://www.odoo.com/submit-pr)

Problem:
The previous fix (replacing '&' with '+' in _replace_characters_SEPA) was applied globally, affecting both reference/identifier fields and human-readable fields such as <Nm> (partner name) and address lines.

As a result, a partner named "test & test GMBH" was exported as:
  <Nm>test + test GMBH</Nm>
instead of the expected:
  <Nm>test & test GMBH</Nm>

This caused bank file rejections because '&' is the correct XML encoding of '&' and is accepted by banks in human-readable fields.

Root cause:
ISO 20022 / EPC217-08 distinguishes two categories of data elements:

- Reference/identifier fields (InstrId, Ustrd, etc.): must use the restricted basic Latin character set — '&' is not allowed and must be replaced with '+'.

- Human-readable fields (Nm, AdrLine, etc.): may contain the extended Latin character set — '&' is valid and must be preserved so lxml can XML-escape it to '&' in the output.

Fix:
Revert the global '&' → '+' replacement in _replace_characters_SEPA so that '&' is preserved for name/address fields. The replacement of '&' with '+' for reference/identifier fields is already handled explicitly at the call sites in _get_CdtTrfTxInf (InstrId, Ustrd) via .replace('&', '+') before sanitize_communication is called.

ref commit : https://github.com/odoo/enterprise/pull/110809/changes/9e698e4ac9fdf66189ff6712f90a144560a1b484
documentation https://www.europeanpaymentscouncil.eu/sites/default/files/KB/files/EPC217-08%20Draft%20Best%20Practices%20SEPA%20Requirements%20for%20Character%20Set%20v1.1.pdf:

Forward-Port-Of: odoo/enterprise#118511
Forward-Port-Of: odoo/enterprise#115409