Numerous `#[test]` functions in `financing_pool_edge_cases.rs` and `access_control_edge_cases.rs` contain no contract calls or assertions, creating false coverage signal

[OxDev-max]

Description

Multiple checked-in #[test] functions consist entirely of comments describing intended
behavior, with no actual contract invocation and no assert!/assert_eq! — meaning they
currently report as "passing" while verifying nothing. In contracts/tests/financing_pool_edge_cases.rs
alone:

• test_yield_calculation_with_equal_positions (lines 83-102)
• test_yield_calculation_with_unequal_positions (lines 104-115)
• test_yield_calculation_with_small_position (lines 117-128)
• test_yield_calculation_with_large_position (lines 130-138)
• test_repayment_lock_prevents_concurrent_repay (lines 197-204)
• test_repayment_lock_cleared_on_success (lines 206-212)
• test_repayment_lock_cleared_on_error (lines 214-220)
• test_repay_exceeds_face_value_allowed (lines 296-302)
• test_repay_already_repaid_fails (lines 317-323)
• test_repayment_completes_when_fully_funded (lines 325-332)
• test_record_position_atomicity (lines 245-266, calls try_record_position but asserts nothing on the result)
• test_repaid_amount_arithmetic_overflow (lines 388-396)

In contracts/tests/access_control_edge_cases.rs:

• test_cannot_transfer_admin_to_zero_address (lines 183-191)
• test_transfer_admin_to_existing_admin_allowed (lines 247-254)

That is 14 confirmed empty/assertion-free tests, all describing genuinely important invariants
(yield-distribution precision, repayment-lock cleanup, over-repayment handling, zero-address
admin-transfer rejection) that are currently not verified at all despite appearing in the
test suite as named, "passing" tests.

Requirements and Context

This is distinct from "add more tests" filler: these tests already exist, are already counted
in CI's green checkmark, and describe exactly what they're supposed to verify in their own
doc comments — they just never call the contract. A reviewer or new contributor skimming test
names would reasonably (and incorrectly) conclude these invariants are covered. Implementing
real bodies requires wiring up actual release_funds/record_position/repay sequences with
real Stellar Asset Contract tokens, mirroring the pattern already proven in
contracts/tests/src/lib.rs's test_multi_investor_partial_recovery_default_end_to_end.

Suggested Execution

1. git checkout -b fix/edge-case-tests-empty-bodies
2. For each of the 12 empty tests in financing_pool_edge_cases.rs, implement a real sequence:
   deploy a mock token, call release_funds/record_position to establish pool state, then
   exercise the specific edge case (equal/unequal/small/large positions; lock-cleared-on-success
   vs. on-error by forcing one call to fail; over-repayment; double-repay rejection; full
   automatic closure) and assert on the resulting Pool/Position/token-balance state.
3. For the 2 empty tests in access_control_edge_cases.rs, either implement a genuine
   zero-address rejection test (Soroban's Address doesn't have a trivial "zero" literal in
   tests, so this may require documenting why it's not directly testable and converting the
   test into a targeted unit test of the underlying validation helper instead), and implement
   real assertions for the idempotent-transfer-to-existing-admin case.
4. Where an edge case turns out to be genuinely untestable in the current test harness (e.g.
   true concurrent reentrancy under Soroban's synchronous model), replace the placeholder with
   an explicit #[ignore] plus a doc comment explaining why, rather than leaving a silently
   passing empty test.
5. Run make fmt and make test.

Acceptance Criteria

• All 12 listed tests in financing_pool_edge_cases.rs contain real contract calls and at least one meaningful assertion
• Both listed tests in access_control_edge_cases.rs contain real assertions or are explicitly marked #[ignore] with a documented reason
• No #[test] function in either file consists solely of comments
• cargo test -p kora-tests shows an increased number of meaningfully-asserting tests, verifiable by diff

Guidelines: PR description must include Closes #<issue-number>.
Complexity: High (200 points)